| *** lyrrad has joined #openstack-swift | 00:00 | |
| *** jasondotstar has quit IRC | 00:01 | |
| *** robefran_ has joined #openstack-swift | 00:04 | |
| *** DericHorn-HP_ has joined #openstack-swift | 00:07 | |
| *** pberis2 has joined #openstack-swift | 00:08 | |
| *** hrou has quit IRC | 00:09 | |
| timur | hey folks, I submitted a bug report a while back and was hoping to get some feedback on it: https://bugs.launchpad.net/swift/+bug/1475018. Do you think it's something that is likely to be fixed or left as Swift quirk? | 00:15 |
|---|---|---|
| openstack | Launchpad bug 1475018 in OpenStack Object Storage (swift) "Either delimiter option is not properly documented or has a bug" [Undecided,New] | 00:15 |
| notmyname | timur: which patch did you submit for it (from the comments there) | 00:16 |
| *** jasondotstar has joined #openstack-swift | 00:16 | |
| *** jasondotstar has quit IRC | 00:16 | |
| *** jasondotstar has joined #openstack-swift | 00:16 | |
| timur | notmyname: https://review.openstack.org/#/c/202302/ | 00:16 |
| *** robefran has quit IRC | 00:17 | |
| *** DericHorn-HP has quit IRC | 00:17 | |
| *** DericHorn-HP_ is now known as DericHorn-HP | 00:17 | |
| timur | most of the patch is a test of the behavior (I noticed that there were no tests for the delimiter behavior -- submitted one more to just add a couple: https://review.openstack.org/#/c/202303/) | 00:18 |
| notmyname | timur: thank you for the tests | 00:20 |
| notmyname | timur: honestly, i don't know yet | 00:21 |
| notmyname | based on the commit message, it may be a bug | 00:21 |
| notmyname | however, "fixing" it is hard depending on how clients are affected | 00:21 |
| timur | notmyname: it's probably not the intended behavior, but I'm not sure if anyone is relying on this quirk | 00:21 |
| timur | right | 00:21 |
| notmyname | yeah | 00:21 |
| notmyname | exactly | 00:22 |
| *** ccavanna has joined #openstack-swift | 00:24 | |
| *** dmorita has joined #openstack-swift | 00:30 | |
| *** darrenc is now known as darrenc_afk | 00:35 | |
| *** lyrrad has quit IRC | 00:40 | |
| ho | :q | 00:43 |
| ho | sorry for it. I typed on wrong app.... | 00:44 |
| *** hrou has joined #openstack-swift | 00:44 | |
| *** DericHorn-HP has quit IRC | 00:54 | |
| *** jasondotstar has quit IRC | 00:56 | |
| *** darrenc_afk is now known as darrenc | 01:00 | |
| *** kutija has quit IRC | 01:07 | |
| ccavanna | Does anyone know about these tests? gate-tempest-dsvm-full | 01:09 |
| ccavanna | I keep getting errors from this test when I check in my code, but I am not sure how to run these tests locally. | 01:10 |
| ccavanna | I'm thinking it's related to this bug. | 01:12 |
| ccavanna | https://bugs.launchpad.net/cinder/+bug/1476735 | 01:12 |
| openstack | Launchpad bug 1476735 in Cinder "check failure in gate-tempest-dsvm-full-ceph; failure to import backup" [High,Confirmed] | 01:12 |
| *** DericHorn-HP has joined #openstack-swift | 01:13 | |
| *** hrou has quit IRC | 01:14 | |
| *** hrou has joined #openstack-swift | 01:15 | |
| openstackgerrit | Carlos Cavanna proposed openstack/swift: Improving statistics sent to Graphite. https://review.openstack.org/202657 | 01:22 |
| ho | ccavanna: i have no experiment to run tempest on my local env but clone https://github.com/openstack/tempest and configure it then execute tempest_run.sh might work. | 01:28 |
| ccavanna | ho: Thanks. I just checked more and there are many reports of that test failing for the past few days. | 01:32 |
| clayg | timur: thanks for asking about a review - i left some thoughts - but it looks great - thanks for the help! | 01:49 |
| *** kota_ has joined #openstack-swift | 01:49 | |
| *** ChanServ sets mode: +v kota_ | 01:49 | |
| kota_ | morning | 01:50 |
| clayg | kota_: !!! | 01:50 |
| clayg | kota_: I feel like you have dozens of patches I haven't looked at | 01:50 |
| clayg | kota_: can I review something for you? | 01:50 |
| kota_ | clayg: please!!!! | 01:50 |
| clayg | kota_: tell me which one | 01:50 |
| clayg | peluse: I didn't get as far as I'd hoped on those tests today | 01:50 |
| clayg | peluse: *but* after torgomatic dropped some knowledge on me I think I have a plan | 01:50 |
| clayg | peluse: just need to make sure no one erased the whiteboard while I was in meetings :'( | 01:51 |
| kota_ | clayg: patch 201055 is highest priority I think | 01:51 |
| patchbot | kota_: https://review.openstack.org/#/c/201055/ | 01:51 |
| kota_ | that's a fix for a bug 1467677 and it is set as critical | 01:51 |
| openstack | bug 1467677 in OpenStack Object Storage (swift) "Server side copy with Single Ranged read not working with Erasure Coded Data" [Critical,In progress] https://launchpad.net/bugs/1467677 - Assigned to Kota Tsuyuzaki (tsuyuzaki-kota) | 01:51 |
| clayg | oh yeah - that looked scary | 01:51 |
| clayg | closes a bug and everything | 01:52 |
| *** DericHorn-HP has quit IRC | 01:52 | |
| kota_ | clayg: for community, I guess patch 199043 should be closed asap. | 01:53 |
| patchbot | kota_: https://review.openstack.org/#/c/199043/ | 01:53 |
| clayg | oh wow - ok so COPY with range param | 01:53 |
| kota_ | clayg: yes, the first one is a fix for COPY with range. | 01:54 |
| kota_ | that causes from the object-server response difference between EC and replicated. | 01:55 |
| clayg | ok, both of those look do-able - thanks! | 01:56 |
| kota_ | on replicated case, we can trust the response status from get_response() call because the range will be already handled on object-server. | 01:56 |
| kota_ | clayg: Thanks for taking care of me :) | 01:57 |
| clayg | lol | 01:57 |
| kota_ | Added a bit, on EC case, the range will be handled on proxy-server, strictly at calling the response (resp()) for starting response. | 01:59 |
| kota_ | that's why the COPY request mishandle the content length and response status and it results in the bug. | 02:00 |
| clayg | sure - that makes sense | 02:00 |
| *** DericHorn-HP has joined #openstack-swift | 02:00 | |
| clayg | kota_: it *does* seem like the copy extraction would have been helpful here | 02:00 |
| clayg | I'm sort of getting tired of "man this would be so much better/easier if copy and versions were just middleware!" | 02:01 |
| kota_ | clayg: exactly, the copy middleware is a solution. do you bet to do so? | 02:02 |
| *** jkugel has joined #openstack-swift | 02:03 | |
| clayg | idk, i just do whatever tdasilva tells me | 02:03 |
| kota_ | lol | 02:03 |
| clayg | i review, find bugs, he fixes them and add tests, I review again | 02:03 |
| clayg | it's a painful process | 02:03 |
| clayg | I assume at some point it will merge and I will stop complaining | 02:03 |
| kota_ | make sense | 02:04 |
| tdasilva | lol | 02:04 |
| *** annegent_ has quit IRC | 02:04 | |
| kota_ | if the copy middleware had depend on just master, I could make an effort to land instead of my patch. | 02:04 |
| notmyname | see the "Big important stuff right now" on https://wiki.openstack.org/wiki/Swift/PriorityReviews | 02:04 |
| kota_ | tdasilva! | 02:05 |
| tdasilva | kota_: let's land versioning first | 02:05 |
| tdasilva | :D | 02:05 |
| notmyname | and encryption needs the copy extration too | 02:05 |
| tdasilva | zaitcev had already given a +2 | 02:05 |
| clayg | versioning hs *got* to be close | 02:05 |
| tdasilva | just needs another one | 02:05 |
| *** annegentle has joined #openstack-swift | 02:06 | |
| tdasilva | clayg: i fixed the latest bug you had found | 02:06 |
| kota_ | ok, I'll be a volunteer | 02:06 |
| tdasilva | yay! | 02:06 |
| clayg | tdasilva: I was sorta pissed when I found that stupid remove bug - I was about to +2 it - but by the time I wrote the functest I was like bah - we might as well go ahead and fix it :'( | 02:06 |
| clayg | tdasilva: yah i was trying to tease out the fix somewhere between patch 35-38 - is x-versions-location the only header that get's set or removed? | 02:07 |
| patchbot | clayg: https://review.openstack.org/#/c/35/ | 02:07 |
| clayg | no patchbot | 02:07 |
| tdasilva | lol | 02:07 |
| kota_ | lol | 02:07 |
| hrou | Hey All, re: vinsh problem from earlier today, I took keystone out of my pipeline (left the OS_ ... set) and can repro: swift list swiftfirstCont --os-storage-url=http://127.0.0.1:8080/v1/AUTH_b2c9f064ec32460eb487aca279a97d76 ValueError: No JSON object could be decoded. So I can reproduce, did we know for a fact keystone was in play ? --debug will show "DEBUG:keystoneclient.auth.identity.v2" if OS_AUTH_ | 02:08 |
| hrou | is set so I don't think that's key | 02:08 |
| clayg | keystone is not in the pipeline and it fails for you? that worked for me! | 02:09 |
| clayg | hrou: thanks for testing btw | 02:09 |
| clayg | you think the environ vars matter? | 02:10 |
| tdasilva | clayg: yeah, when you send x-remove-versions-location, it just sets x-versions-location to '' | 02:10 |
| *** annegentle has quit IRC | 02:10 | |
| hrou | clayg, did you use curl or the swift client ? Seems like the problem lies there, I can use 'curl' just fine and get the html listing. | 02:10 |
| clayg | tdasilva: that looks pretty sane to me! | 02:11 |
| tdasilva | kota_: are you going to be in Austin? | 02:11 |
| clayg | kota_: say yes say yes! | 02:11 |
| clayg | tdasilva: are YOU!? | 02:11 |
| kota_ | yes! | 02:12 |
| clayg | kota_: YES! | 02:12 |
| tdasilva | cool, i'll be there too | 02:12 |
| tdasilva | looking forward to seeing you guys there | 02:12 |
| kota_ | great :D | 02:12 |
| clayg | hrou: yeah with curl (no auth token) -> html, with curl (x-autht-token) -> json, with cli (x-auth-token) -> json | 02:13 |
| kota_ | i'm now writing a new spec for swift, maybe i could introduce it at Austin. | 02:13 |
| tdasilva | kota_: cool...what you working on? | 02:13 |
| kota_ | I'm not sure it would be good. | 02:13 |
| clayg | kota_: LOL | 02:13 |
| kota_ | working on EC efficiation for global cluster. | 02:14 |
| *** DericHorn-HP has quit IRC | 02:14 | |
| kota_ | strictly 2 data center EC. | 02:14 |
| kota_ | sometimes custamer wants to ensure the saving objects if one data center gone also on EC case. | 02:15 |
| kota_ | It absolutely needs equal or more than 2x data redanduncy but... | 02:16 |
| kota_ | 2x data (e.g. k=20, m=24) redanduncy decreases encoding/decoding peforemance sinificantly. | 02:17 |
| kota_ | so i'm now thinking/writing a new way changing Swift handing instead of using such a large number of parities. | 02:19 |
| clayg | kota_: I always assumed that multi-site EC for these DR scenarios would just want two replicas of the EC'd data - 10+4 over here and 6+2 over there. | 02:21 |
| kota_ | According to spec process, I have to write idea and goals w/o patch design but I must think the design because I am a developer :P | 02:21 |
| clayg | kota_: start with the use case, and design from the cluster operator up | 02:21 |
| kota_ | clayg: yes, my idea is similar with yours. | 02:22 |
| clayg | kota_: neat! | 02:22 |
| kota_ | yes, start with the use case, I really agree! | 02:22 |
| clayg | :D | 02:22 |
| *** blmartin has joined #openstack-swift | 02:23 | |
| clayg | hopefully some of the IBM/SoftLayer guys will be in austin - I think for some deployments a better container-sync implementation will be be exactly what the deployer needs | 02:24 |
| clayg | does anyone remember or has seen the per-disk-rsync-module patch? | 02:24 |
| *** gyee has quit IRC | 02:39 | |
| *** zaitcev has joined #openstack-swift | 02:39 | |
| *** ChanServ sets mode: +v zaitcev | 02:39 | |
| kota_ | wow, versions_write middleware has amazing number of patch sets. | 02:50 |
| clayg | tdasilva: is amazing | 02:50 |
| kota_ | absolutely | 02:51 |
| *** peterbus_ has joined #openstack-swift | 02:53 | |
| *** blmartin has quit IRC | 02:53 | |
| *** DericHorn-HP has joined #openstack-swift | 02:57 | |
| *** DericHorn-HP has quit IRC | 02:58 | |
| *** nakagawamsa has joined #openstack-swift | 03:01 | |
| *** pberis2 has quit IRC | 03:01 | |
| *** DericHorn-HP has joined #openstack-swift | 03:02 | |
| *** DericHorn-HP has quit IRC | 03:13 | |
| *** sanchitmalhotra has joined #openstack-swift | 03:17 | |
| openstackgerrit | Azhagu Selvan SP proposed openstack/swift: Respect 'Accept' header in error responses https://review.openstack.org/204196 | 03:31 |
| *** h00327910_ has quit IRC | 03:35 | |
| *** DericHorn-HP has joined #openstack-swift | 03:36 | |
| *** sanchitmalhotra has quit IRC | 03:36 | |
| openstackgerrit | Azhagu Selvan SP proposed openstack/swift: Respect 'Accept' header in error responses https://review.openstack.org/204196 | 03:36 |
| *** sanchitmalhotra has joined #openstack-swift | 03:37 | |
| *** km_ has joined #openstack-swift | 04:03 | |
| *** km has quit IRC | 04:03 | |
| *** kei_yama has quit IRC | 04:04 | |
| *** kei_yama has joined #openstack-swift | 04:05 | |
| *** kota_ has quit IRC | 04:06 | |
| *** ho_ has joined #openstack-swift | 04:07 | |
| *** ho has quit IRC | 04:09 | |
| *** Kennan2 has joined #openstack-swift | 04:14 | |
| *** Kennan has quit IRC | 04:15 | |
| *** DericHorn-HP has quit IRC | 04:35 | |
| *** kota_ has joined #openstack-swift | 04:41 | |
| *** ChanServ sets mode: +v kota_ | 04:41 | |
| *** ccavanna has quit IRC | 04:45 | |
| zaitcev | Does anyone remember if policy names permit special characters such as dots or colons? | 04:50 |
| *** tsg has quit IRC | 04:52 | |
| clayg | zaitcev: we started really restrictive | 04:53 |
| clayg | I think you only get stuff that will work in a header key | 04:54 |
| clayg | so like ascii letters numbers and hyphen's | 04:54 |
| clayg | I don't think you can even have underscores? | 04:54 |
| clayg | it's in storage_policy.py IIRC | 04:54 |
| kota_ | valid characters definitions is here? VALID_CHARS = '-' + string.letters + string.digits | 04:55 |
| zaitcev | kota_, clayg: I see, thanks. | 04:56 |
| kota_ | clayg: thanks for reviewing patch 199043 :) | 04:56 |
| patchbot | kota_: https://review.openstack.org/#/c/199043/ | 04:56 |
| clayg | kota_: that shit is crazy | 04:57 |
| clayg | i'm almost done with the EC Copy Range test | 04:57 |
| clayg | i don't quite yet understand why it's not just better to wait and set content_length later | 04:57 |
| *** darrenc is now known as darrenc_afk | 04:58 | |
| kota_ | to wait? | 04:59 |
| clayg | well that iter has to get kicked eventually - but I guess it's not until after we've already established backend connections? | 04:59 |
| *** SkyRocknRoll has joined #openstack-swift | 05:00 | |
| clayg | yup ok | 05:00 |
| kota_ | yup, there are 2 different iter in the EC resposne, one is range iter and the other is body iter. | 05:02 |
| *** kutija has joined #openstack-swift | 05:03 | |
| kota_ | Normaly, if we need a single range, ResumingGetter will knock the unecessary iter but | 05:05 |
| *** ho has joined #openstack-swift | 05:05 | |
| kota_ | it doesn't affect the backend resuming getter. | 05:05 |
| clayg | are we talking about patch 201055 or the other one I already reviewed? | 05:05 |
| patchbot | clayg: https://review.openstack.org/#/c/201055/ | 05:05 |
| kota_ | yes, | 05:06 |
| clayg | 499 copy range | 05:06 |
| kota_ | ok, sorry. | 05:06 |
| clayg | and the range iter matters here? | 05:06 |
| *** kei_yama has quit IRC | 05:06 | |
| *** km_ has quit IRC | 05:06 | |
| kota_ | no relation between 199043 and 201055, they are isolated. | 05:07 |
| *** ho_ has quit IRC | 05:07 | |
| clayg | ok - so which one are we talking about - i'm working on 499 copy range - every time I think I have it; it slips | 05:07 |
| *** ho has quit IRC | 05:07 | |
| kota_ | ok, I was misunderstanding you might talk about 199043 | 05:08 |
| kota_ | that's my fault. | 05:08 |
| clayg | the tests are great tho, and change *looks* simple - but it's down in the guts of swob and there's all these iters wrapping iters - it just hurts | 05:08 |
| clayg | oh | 05:08 |
| clayg | yeah we can talk about that one too | 05:08 |
| clayg | the two client disconnect messages on master are two ec_ndata connections | 05:08 |
| clayg | I don't see how the range iters really play ... oh you're saying *thats* what's raising the Generater Exit? | 05:09 |
| *** kei_yama has joined #openstack-swift | 05:09 | |
| kota_ | the Generater Exit raising is about 199043. | 05:10 |
| clayg | yeah ok | 05:11 |
| clayg | so patch 199043 | 05:11 |
| patchbot | clayg: https://review.openstack.org/#/c/199043/ | 05:11 |
| kota_ | actually, the copy will succeed w/o 201055 but proxy handles the succeeded put as 499 client disconnected | 05:11 |
| *** ppai has joined #openstack-swift | 05:11 | |
| clayg | see now *you're* mixing the two | 05:11 |
| clayg | oh wait... maybe you're just talking about 201055 | 05:11 |
| clayg | which one do you want to talk about :D | 05:12 |
| kota_ | yes, for now I'm talking about 201055. | 05:12 |
| clayg | I was done with 199043 for the moment unless you wanted to drop some knowledge on me - I *did* -1 it - so you may need to strighten me out | 05:12 |
| kota_ | on the 201055 case, no GeneratorExit happned. | 05:12 |
| clayg | I think I like 201055, but I'm still struggling | 05:12 |
| clayg | ok yeah so 201055 | 05:12 |
| clayg | that's 499 copy range | 05:13 |
| clayg | k | 05:13 |
| kota_ | for example | 05:13 |
| kota_ | the source object size is 100 bytes | 05:13 |
| clayg | I think I don't understand why it worked on the replicated case | 05:13 |
| kota_ | and then wants to copy the object with range:bytes=5-10 | 05:14 |
| kota_ | and ec segment size is 1024 | 05:14 |
| kota_ | and k=2, m=1, so assuming actual fragment size 512 bytes. | 05:15 |
| kota_ | proxy server will get whole bytes (i.e. 512 bytes) from object-server to decode the original segment. | 05:16 |
| kota_ | the response from src_req.get_response() will be 200 and content_length will be 512. | 05:17 |
| kota_ | proxy sever will set the 512 as target_req.content_length. | 05:17 |
| kota_ | but actuall transfer byte size is just 5 bytes because of range header. | 05:18 |
| clayg | yeah i get all that | 05:18 |
| kota_ | it will make 499 client disconnected | 05:18 |
| kota_ | thanks | 05:18 |
| clayg | just trying to understand the change - the bug makes perfect sense | 05:18 |
| kota_ | oh, ok | 05:19 |
| clayg | so before EC did swob not used to do any of this content_length manipulation? | 05:19 |
| kota_ | In replicated case, the object-server response will be 206 partial conetn and src_resp.content_length will be 5. | 05:19 |
| clayg | yup - see that now | 05:20 |
| kota_ | so in replicated case, we don't have to consider about that on proxy, just transferring was enough. | 05:21 |
| *** zaitcev has quit IRC | 05:21 | |
| clayg | kota_: yes but *why* is calling _respose_iter the *right* fix? | 05:22 |
| clayg | i understand the bug and I've duplicated the code change resolves the issue | 05:22 |
| clayg | but why is this fix the right fix | 05:22 |
| clayg | "it works" is rarely good enough - _respose_iter is a under method - swob didn't expect you to need this | 05:22 |
| kota_ | oic... | 05:23 |
| clayg | I see that it fixes the content_length but why does it have to wait to do that in *that* method? | 05:23 |
| clayg | why doesn't the source_resp that comes out of self.GET just have the right content-lenght on it | 05:24 |
| *** hrou has quit IRC | 05:25 | |
| kota_ | so you means the _reponse_iter will be called, we should wait to do that. | 05:25 |
| clayg | i don't really know - that's what I'm struggling with - it's quite possible this is the perfect fix - i just don't understand why | 05:26 |
| kota_ | i guess we should take the right content_length before opening the backend target connection. | 05:26 |
| clayg | oh | 05:26 |
| clayg | ECObjectController._fix_response_headers I think is not so correct about 'Content-Length' | 05:27 |
| *** ho has joined #openstack-swift | 05:27 | |
| kota_ | or we should use just encoding: chunked without content length? | 05:27 |
| clayg | I prefer to keep all the extra cunked bytes off the wire when we can | 05:28 |
| kota_ | clayg: yup, _fix_reponse_headers just set original conent length. | 05:28 |
| clayg | kota_: so but that's stupid right? | 05:28 |
| kota_ | clayg: | 05:28 |
| kota_ | right | 05:28 |
| kota_ | but once i changed that, swift doesn't work well. | 05:29 |
| clayg | wtf - how can it even do that? what about a 404 response - or 503 | 05:29 |
| clayg | kota_: lol! | 05:29 |
| kota_ | the content length set by _fix_response_headers seems to be referenced | 05:30 |
| kota_ | maybe range calculation? | 05:30 |
| clayg | oh i bet it does that but then swob fixes it | 05:30 |
| clayg | :P | 05:30 |
| clayg | because if the body is a static html it can just call length | 05:30 |
| clayg | and we already know that ranged requests will fix content length | 05:30 |
| kota_ | i see | 05:31 |
| clayg | goodness gracious | 05:31 |
| *** mahatic has joined #openstack-swift | 05:41 | |
| openstackgerrit | Matthew Oliver proposed openstack/swift: Add container and account reverse listings https://review.openstack.org/120709 | 05:43 |
| kota_ | wait, src_response would *not* call _response_iter because the response doesn't be passed back to __call__, right? | 05:48 |
| kota_ | handle_copy_request just uses the app_iter as just a data source like as wsgi.input | 05:49 |
| kota_ | i might have to re-read swob implementation... | 05:50 |
| clayg | kota_: yeah thats where I am | 05:50 |
| clayg | kota_: but yeah I don't think the replicated case ever gets into this method - the content length is just right because the object server knows how to set it | 05:51 |
| clayg | I'm pretty sure the GET method on the ECObjectController should do the same - but ultimately it's swob job to do all that range parsing bits | 05:51 |
| kota_ | exactly | 05:51 |
| clayg | it's just got it self setup to not do it until the last minute | 05:51 |
| clayg | although it *could* do it anytime after you set the body/app_iter and content_length to the real obj size | 05:52 |
| clayg | you could do it in __init__ but the property setter interface lets you change .content_length after you __init__ with an app_iter :\ | 05:52 |
| kota_ | oh, sorry I have a meeting since 5 minutes later. I'm willing to make sure how we could improve this later. | 05:53 |
| kota_ | thanks clayg! | 05:53 |
| *** ktsuyuzaki has joined #openstack-swift | 05:55 | |
| *** ktsuyuzaki has quit IRC | 05:56 | |
| *** kota_ has quit IRC | 05:58 | |
| *** mahatic_ has joined #openstack-swift | 06:03 | |
| *** mahatic has quit IRC | 06:04 | |
| *** darrenc_afk is now known as darrenc | 06:05 | |
| *** mahatic has joined #openstack-swift | 06:06 | |
| *** mahatic_ has quit IRC | 06:07 | |
| *** mahatic has quit IRC | 06:14 | |
| jith_ | hi all, in dashboard->images(glance) what is the architecture field while creating images | 06:16 |
| *** silor has joined #openstack-swift | 06:19 | |
| clayg | jith_: did you try in #openstack-glance already? | 06:20 |
| clayg | jith_: you might try in #openstck-dev too | 06:20 |
| *** silor1 has joined #openstack-swift | 06:20 | |
| jith_ | clayg: ya i tried in glance.. will try in dev too.. thanks | 06:20 |
| clayg | it's good to try here - some swift people know people that use glance ;) | 06:21 |
| *** silor has quit IRC | 06:24 | |
| *** silor1 is now known as silor | 06:24 | |
| *** mahatic has joined #openstack-swift | 06:24 | |
| jith_ | yes.... any one configured swift as backend for glance?? | 06:25 |
| *** ppai has quit IRC | 06:27 | |
| *** mahatic has quit IRC | 06:32 | |
| *** mahatic has joined #openstack-swift | 06:33 | |
| *** ppai has joined #openstack-swift | 06:39 | |
| *** mfalatic has quit IRC | 06:42 | |
| *** mahatic has quit IRC | 06:52 | |
| *** lyrrad has joined #openstack-swift | 06:53 | |
| clayg | kota_ there's you go - two -1's - last time you ask me for reviews | 06:56 |
| *** hseipp has joined #openstack-swift | 07:04 | |
| mattoliverau | clayg: lol, look at you commenting on a commit message in patch 202303, I'm so proud :P If you want to push up a new patch with a better commit message I'd be happy to +2 it ;) | 07:05 |
| patchbot | mattoliverau: https://review.openstack.org/#/c/202303/ | 07:05 |
| clayg | it's just free tests - timur asked in channel so I think he was looking for the feedback - we'll see what he says | 07:05 |
| clayg | I do sorta honestly remember one time thinking that a prefix query with a delim, but when the prefix didn't *end* in the delim was basically non-sensical | 07:06 |
| clayg | but i guess that's not really one of the tests | 07:07 |
| mattoliverau | yeah, I was just battling prefixes + delimiters in the reverse listing patch, so found it funny to open up a review to some testing of the same kinda stuff. | 07:09 |
| clayg | lol - right! | 07:09 |
| clayg | you probably could have written it better! | 07:09 |
| mattoliverau | clayg: so your back from holidays, and you do a late nighter. Great to have you back! :) | 07:10 |
| clayg | thanks! | 07:11 |
| *** mahatic has joined #openstack-swift | 07:17 | |
| *** kota_ has joined #openstack-swift | 07:20 | |
| *** ChanServ sets mode: +v kota_ | 07:20 | |
| *** ktsuyuzaki has joined #openstack-swift | 07:23 | |
| ktsuyuzaki | wow, my wi-fi at meeting room is not well... | 07:24 |
| ktsuyuzaki | this is kota_ | 07:24 |
| -openstackstatus- NOTICE: zuul is undergoing an emergency restart. Jobs will be re-queued but some events may be lost. | 07:24 | |
| openstackgerrit | Kota Tsuyuzaki proposed openstack/swift: Fix 499 client disconnected on COPY EC object https://review.openstack.org/201055 | 07:25 |
| *** lyrrad has quit IRC | 07:25 | |
| *** kota_ has quit IRC | 07:26 | |
| *** kota_ has joined #openstack-swift | 07:29 | |
| *** ChanServ sets mode: +v kota_ | 07:29 | |
| clayg | kota_: I think that's already looking better | 07:31 |
| *** ktsuyuzaki has quit IRC | 07:31 | |
| clayg | kota_: if you're still working on it I'll hang back - but I think the direction that you're taking there fits pretty well with my understanding of the problem | 07:32 |
| clayg | kota_: I think it's worth nothing fix_response_headers will also fix the status - which is important for the error handling in the copy logic - because it checks for 4XX on the GET part and returns the immediately - which is key | 07:34 |
| *** ktsuyuzaki has joined #openstack-swift | 07:34 | |
| ktsuyuzaki | clayg: thanks, in my quick view, I also think my direction addressed yours so just let me have a time to review and to think. I'll ping you during today on gerrit, again after I confirmed the diff more deeply. | 07:35 |
| *** kota_ has quit IRC | 07:36 | |
| clayg | i like your diff better than mine <blush> | 07:37 |
| clayg | but np, just ping me when it it needs another round | 07:38 |
| openstackgerrit | Hisashi Osanai proposed openstack/swift: Add functional test for access control (PUT) with Keystone https://review.openstack.org/202411 | 07:39 |
| *** jordanP has joined #openstack-swift | 07:39 | |
| ktsuyuzaki | clayg: perhaps, you hit another bug 1475499 during testing the range copy | 07:41 |
| openstack | bug 1475499 in OpenStack Object Storage (swift) "EC: proxy server returns wrong response on range GET" [Undecided,Confirmed] https://launchpad.net/bugs/1475499 - Assigned to Daisuke Morita (morita-daisuke) | 07:41 |
| ktsuyuzaki | clayg: the first hang curl on https://gist.github.com/clayg/cf1c39be98279ae3d7eb | 07:41 |
| clayg | ktsuyuzaki: heh - yeah that one "hung" waiting for the 62442704 bytes that would never come ;) | 07:42 |
| ktsuyuzaki | i guess that's not only COPY | 07:42 |
| ktsuyuzaki | also affects GET | 07:42 |
| clayg | oh no? any unsatisfiable range on GET | 07:42 |
| clayg | I tried it on master and I didn't think it hung like that? | 07:43 |
| clayg | for... some reason? | 07:43 |
| ktsuyuzaki | really? for EC object? | 07:43 |
| clayg | oh... maybe it totally do.. no that doesn't make sense | 07:43 |
| clayg | bha yeah - that's lame | 07:44 |
| *** ppai has quit IRC | 07:46 | |
| openstackgerrit | Timur Alperovich proposed openstack/swift: Add two functional tests for delimiter. https://review.openstack.org/202303 | 07:47 |
| *** kota_ has joined #openstack-swift | 07:49 | |
| *** ChanServ sets mode: +v kota_ | 07:49 | |
| openstackgerrit | Timur Alperovich proposed openstack/swift: Add two functional tests for delimiter. https://review.openstack.org/202303 | 07:50 |
| *** ktsuyuzaki has quit IRC | 07:51 | |
| *** ktsuyuzaki has joined #openstack-swift | 07:55 | |
| *** kota_ has quit IRC | 07:57 | |
| *** kota_ has joined #openstack-swift | 07:58 | |
| *** ChanServ sets mode: +v kota_ | 07:58 | |
| openstackgerrit | Hisashi Osanai proposed openstack/swift: Add functional test for access control (DELETE) with Keystone https://review.openstack.org/202986 | 07:58 |
| *** ppai has joined #openstack-swift | 07:59 | |
| *** ktsuyuzaki has quit IRC | 08:00 | |
| openstackgerrit | Hisashi Osanai proposed openstack/swift: Add functional test for access control (GET) with Keystone https://review.openstack.org/202994 | 08:01 |
| *** ktsuyuzaki has joined #openstack-swift | 08:02 | |
| *** jistr has joined #openstack-swift | 08:03 | |
| openstackgerrit | Hisashi Osanai proposed openstack/swift: Add functional test for access control (HEAD) with Keystone https://review.openstack.org/202997 | 08:04 |
| *** kota_ has quit IRC | 08:04 | |
| openstackgerrit | Clay Gerrard proposed openstack/swift: wip: tests for paul https://review.openstack.org/204485 | 08:06 |
| openstackgerrit | Hisashi Osanai proposed openstack/swift: Add functional test for access control (POST) with Keystone https://review.openstack.org/202999 | 08:07 |
| openstackgerrit | Hisashi Osanai proposed openstack/swift: Add functional test for access control (OPTIONS) with Keystone https://review.openstack.org/203012 | 08:09 |
| *** aix has joined #openstack-swift | 08:10 | |
| *** ktsuyuzaki has quit IRC | 08:11 | |
| ho | jith_: fyi. http://docs.openstack.org/cli-reference/content/chapter_cli-glance-property.html | 08:11 |
| *** kota_ has joined #openstack-swift | 08:15 | |
| *** ChanServ sets mode: +v kota_ | 08:15 | |
| jith_ | ho: thanks :) | 08:16 |
| -openstackstatus- NOTICE: zuul jobs are currently stuck while problems with gearman are debugged | 08:16 | |
| *** ChanServ changes topic to "zuul jobs are currently stuck while problems with gearman are debugged" | 08:16 | |
| openstackgerrit | Clay Gerrard proposed openstack/swift: do container listing updates in another (green)thread https://review.openstack.org/189080 | 08:20 |
| clayg | so that's it for me i think | 08:20 |
| clayg | dose *anyone* remember the per-disk-rsync module patch? Did I dream it? | 08:21 |
| ho | clayg: how about patch 192186 ? | 08:24 |
| patchbot | ho: https://review.openstack.org/#/c/192186/ | 08:24 |
| clayg | ho: yes! | 08:25 |
| ho | clayg: it's not your dream :) | 08:26 |
| kota_ | clayg: http://paste.openstack.org/show/399105/ would resolve your curl stuck on https://gist.github.com/clayg/cf1c39be98279ae3d7eb | 08:29 |
| clayg | yeah I must have been doing some testing wih that change in place | 08:29 |
| kota_ | clayg: we need absolutely more improvement | 08:30 |
| clayg | I think I like the trick where you add the resp.new_swob_method_that_makes_conditional_responses_not_stupid() into self._fix_response_headers too | 08:30 |
| clayg | patch it! | 08:30 |
| clayg | go go go kota! | 08:30 |
| clayg | mattoliverau can plus +2 it and I'll +A in the am | 08:30 |
| clayg | go world team swift! | 08:31 |
| kota_ | k, i wanna make sure you mean new_swob_method_that_makes_conditional_responses_not_stupid()...sorry. | 08:33 |
| kota_ | should i change the method name on swob on current patch? | 08:35 |
| ho | acoles_wa | 08:35 |
| ho | acoles_away, donagh: around? | 08:35 |
| clayg | what!? | 08:37 |
| clayg | kota_: what did you name it | 08:37 |
| kota_ | swob.Response.fix_response_headers | 08:37 |
| clayg | it was like fix_response_headers - which was close - but it should also capture the idea that it fixes the response status too and whatever "conditional_range" is | 08:37 |
| *** joeljwright has joined #openstack-swift | 08:38 | |
| *** ChanServ sets mode: +v joeljwright | 08:38 | |
| kota_ | ok | 08:38 |
| clayg | I think it will also call close_if_possilbe(app_iter) on error - so - it does a lot | 08:38 |
| *** kei_yama has quit IRC | 08:38 | |
| clayg | but no, "new_swob_method_that_makes_conditional_responses_not_stupid" while an accurate name - was ment to be a joke and a placeholder | 08:38 |
| clayg | it's a bit long to type | 08:39 |
| kota_ | k | 08:39 |
| kota_ | maybe just "fix_conditional_response" seems good from your patch on github. | 08:40 |
| clayg | +1 | 08:40 |
| clayg | mattoliverau: didn't you really hate on us for the non-dry ecdiskfile stuff - go review https://review.openstack.org/#/c/198429/ already! | 08:42 |
| *** rledisez has joined #openstack-swift | 08:48 | |
| *** ppai has quit IRC | 08:55 | |
| openstackgerrit | Kota Tsuyuzaki proposed openstack/swift: Fix 499 client disconnected on COPY EC object https://review.openstack.org/201055 | 08:59 |
| kota_ | oh I forgot rewrite commit log... | 09:01 |
| ho | acoles_away, donagh: as we agreed in last summit, I have been developing functional test for access control. I try to split into small patches so I would like to have reviews on patch 202411 first. I will submit several patches on it within two weeks. | 09:06 |
| patchbot | ho: https://review.openstack.org/#/c/202411/ | 09:06 |
| clayg | whoa - that's so meta | 09:07 |
| clayg | ho: http://stackoverflow.com/questions/6453235/what-does-damp-not-dry-mean-when-talking-about-unit-tests | 09:08 |
| *** ppai has joined #openstack-swift | 09:08 | |
| clayg | ho: not to say there's anything wrong with those tests - but first impression -> xac means nothing to me, and that RBAC_PUT table doesn't immediately provide a lot of insight into the expected behavior of the UUT | 09:10 |
| openstackgerrit | Kota Tsuyuzaki proposed openstack/swift: Fix 499 client disconnected on COPY EC object https://review.openstack.org/201055 | 09:10 |
| clayg | ho: don't listen to me - it's late - you know what they say the difference in good code vs bad code? http://ptrace.fefe.de/wtfm.jpg | 09:11 |
| clayg | i'm sure it'll make sense when I dig in ;) | 09:11 |
| clayg | g'night all | 09:11 |
| kota_ | done to fix some method name. | 09:11 |
| kota_ | and commit log | 09:12 |
| *** km has joined #openstack-swift | 09:24 | |
| *** nakagawamsa has quit IRC | 09:27 | |
| *** jasondotstar has joined #openstack-swift | 09:30 | |
| *** jasondotstar has quit IRC | 09:35 | |
| *** ppai has quit IRC | 09:45 | |
| *** ppai has joined #openstack-swift | 09:59 | |
| *** joeljwright1 has joined #openstack-swift | 10:07 | |
| ho | clayg: thanks for the info. unnn... the style of the functional test was discussed in the summit ("policy.json" in log: https://etherpad.openstack.org/p/liberty-swift-contributors-meetup). the tests should be expressed as a matrix in the code and a line consists of condtions and an expected result in order to validate current access control behaviors and easy to understand expectation. | 10:07 |
| ho | clayg: I almost finished additional functional test for container acl now with same style... So I would like to discuss it tomorrow. | 10:07 |
| *** joeljwright2 has joined #openstack-swift | 10:08 | |
| *** joeljwright has quit IRC | 10:10 | |
| *** joeljwright1 has quit IRC | 10:11 | |
| ho | clayg: good night! | 10:11 |
| *** SkyRocknRoll has quit IRC | 10:17 | |
| *** aix has quit IRC | 10:19 | |
| *** DericHorn-HP has joined #openstack-swift | 10:20 | |
| *** kota_ has quit IRC | 10:25 | |
| *** joeljwright2 has quit IRC | 10:46 | |
| *** jasondotstar has joined #openstack-swift | 10:48 | |
| *** robefran_ has quit IRC | 10:54 | |
| *** ppai has quit IRC | 10:58 | |
| *** joeljwright has joined #openstack-swift | 11:02 | |
| *** ChanServ sets mode: +v joeljwright | 11:02 | |
| *** joeljwright1 has joined #openstack-swift | 11:05 | |
| *** SkyRocknRoll has joined #openstack-swift | 11:05 | |
| *** joeljwright has quit IRC | 11:07 | |
| *** SkyRocknRoll has quit IRC | 11:08 | |
| *** mahatic has quit IRC | 11:10 | |
| *** ppai has joined #openstack-swift | 11:11 | |
| *** mahatic has joined #openstack-swift | 11:12 | |
| *** aix has joined #openstack-swift | 11:18 | |
| *** jkugel has quit IRC | 11:21 | |
| *** SkyRocknRoll has joined #openstack-swift | 11:25 | |
| *** DericHorn-HP has quit IRC | 11:30 | |
| *** ho has quit IRC | 11:36 | |
| *** mahatic has quit IRC | 11:43 | |
| openstackgerrit | Ondrej Novy proposed openstack/swift: Rsync port in ring file. https://review.openstack.org/192198 | 11:49 |
| *** SkyRocknRoll has quit IRC | 11:52 | |
| *** david-lyle has joined #openstack-swift | 11:52 | |
| onovy | clayg, hi clay. if you have any more questions for 192198, i'm here to reply | 11:54 |
| mattoliverau | onovy: I think he's gone to bed, it is very late (or is it very early now) in the states. | 11:55 |
| onovy | ah, ok, thanks | 11:55 |
| onovy | how long takes code review? avg? :) i'm waiting 1mon+, and almost nothing happends :( https://review.openstack.org/#/c/192562/ | 11:58 |
| *** mahatic has joined #openstack-swift | 11:59 | |
| mattoliverau | onovy: good question, depends on what's happening in the community and time of year. Or if we are frantically trying to get a major feature in the next release. | 12:02 |
| mattoliverau | Sorry if sometimes it takes a while. | 12:02 |
| onovy | np, is there something what can i do for review? | 12:02 |
| *** sanchitmalhotra has quit IRC | 12:04 | |
| mattoliverau | onovy: Your doing the right thing, ask in here if it seems stalled. I'm about to goto bed (it's 10pm in Australia), but will add your change to the top of my list tomorrow :) | 12:04 |
| onovy | thanks a lot, good night | 12:05 |
| mattoliverau | night :) | 12:05 |
| *** peterbus_ has quit IRC | 12:19 | |
| *** peterbusque has joined #openstack-swift | 12:20 | |
| openstackgerrit | Christian Schwede proposed openstack/swift: Removing commented out code in test/unit/account/test_backend.py https://review.openstack.org/204575 | 12:24 |
| *** dmorita has quit IRC | 12:26 | |
| *** mfalatic has joined #openstack-swift | 12:27 | |
| *** tongli has joined #openstack-swift | 12:30 | |
| *** ppai has quit IRC | 12:31 | |
| *** km has quit IRC | 12:40 | |
| *** DericHorn-HP has joined #openstack-swift | 12:44 | |
| *** hrou has joined #openstack-swift | 12:54 | |
| *** annegentle has joined #openstack-swift | 12:56 | |
| *** haypo has joined #openstack-swift | 12:57 | |
| haypo | cschwede: hi. i replied to your comments on https://review.openstack.org/#/c/199034/ | 12:57 |
| haypo | cschwede: this patch is the shortest patch to reach the miletone "working tox -e py34" which will be a nice step to avoid Python 3 regressions on the tested code | 12:58 |
| *** joeljwright1 has quit IRC | 12:58 | |
| *** minwoob has joined #openstack-swift | 12:59 | |
| *** joeljwright has joined #openstack-swift | 13:02 | |
| *** ChanServ sets mode: +v joeljwright | 13:02 | |
| cschwede | haypo: well, the change looks mostly ok to me, except the rfc822 workaround and pyeclib workaround in tox.ini. but looks like both issues are making progress, which is a good thing | 13:09 |
| openstackgerrit | paul luse proposed openstack/swift: Add ability for GET path to see/select alternate frag archs https://review.openstack.org/201283 | 13:10 |
| *** robefran has joined #openstack-swift | 13:15 | |
| *** DericHorn-HP has quit IRC | 13:19 | |
| *** jkugel has joined #openstack-swift | 13:19 | |
| *** SkyRocknRoll has joined #openstack-swift | 13:21 | |
| *** david-lyle has quit IRC | 13:27 | |
| *** wer has joined #openstack-swift | 13:28 | |
| onovy | https://review.openstack.org/#/c/192562/4/AUTHORS should i remove my name, as paul suggests and resend patch? | 13:48 |
| *** wbhuber has joined #openstack-swift | 13:51 | |
| *** DericHorn-HP has joined #openstack-swift | 13:58 | |
| *** blmartin has joined #openstack-swift | 14:00 | |
| *** jrichli has joined #openstack-swift | 14:02 | |
| *** tsg has joined #openstack-swift | 14:07 | |
| wbhuber | peluse: what's involved in the tuned that tweaks settings for incremental performance? code changes within diskfile? even though the charts are straightforward, but what's involved between "untuned" and "tuned" remains ambiguous... | 14:12 |
| *** SkyRocknRoll has quit IRC | 14:12 | |
| *** SkyRocknRoll has joined #openstack-swift | 14:13 | |
| mahatic | jrichli, hi! I see that you have unit tests (for encrypter on PUT, POST etc and missing keys as well). Could you tell me what probe tests are supposed to be verifying? | 14:15 |
| -openstackstatus- NOTICE: CI is currently recovering from an outage overnight. It is safe to recheck results with NOT_REGISTERED errors. It may take some time for zuul to work through the backlog. | 14:16 | |
| *** ChanServ changes topic to "CI is currently recovering from an outage overnight. It is safe to recheck results with NOT_REGISTERED errors. It may take some time for zuul to work through the backlog." | 14:16 | |
| jrichli | mahatic - sure. There are two cards in trello for the probetests. One of them is: | 14:17 |
| jrichli | Probetest: PUT object, verify stored object is encrypted, verify container listing metadata is consistent and encrypted with container key | 14:17 |
| jrichli | The first step is for you to setup your SAIO that includes the latest for encryption. | 14:18 |
| mahatic | jrichli, hmm, I did look at them. But aren't you already verifying "Stored object is encrypted" in your unit test? | 14:19 |
| jrichli | to do this, download the code from https://review.openstack.org/#/c/203454/ | 14:19 |
| jrichli | tbh, acoles_away wrote that card. I think that the purpose is to have an integrated test with the real encryption/decryption middleware and the crypto class. | 14:20 |
| mahatic | okay, that doesn't sound too straightforward | 14:25 |
| *** haomaiwang has joined #openstack-swift | 14:25 | |
| jrichli | I am happy to help with any questions you have. | 14:28 |
| mahatic | yes sure, I was looking at the code | 14:28 |
| *** maestro has joined #openstack-swift | 14:28 | |
| maestro | hi | 14:28 |
| maestro | i want to increase the speed download | 14:29 |
| maestro | in swift | 14:29 |
| maestro | what are the parameters to consider to do fine tune | 14:29 |
| maestro | presently am getting 75Mbps.. but still want to increase | 14:30 |
| maestro | 75MBps | 14:30 |
| jrichli | mahatic: be sure not to look at the old review anymore. Are you looking at what is there when you download patch 203454? | 14:32 |
| patchbot | jrichli: https://review.openstack.org/#/c/203454/ | 14:32 |
| mahatic | jrichli, I didn't download that yet, I'm still looking at it directly over the link | 14:33 |
| *** ChanServ changes topic to "Review Dashboard: http://goo.gl/8IUcKl | Summary Dashboard: http://goo.gl/qHus5v | Hackathon topics: https://etherpad.openstack.org/p/swift-midcycle-aug-2015 | Logs: http://eavesdrop.openstack.org/irclogs/%23openstack-swift/" | 14:35 | |
| -openstackstatus- NOTICE: CI is slowly recovering, please be patient while the backlog is worked through. | 14:35 | |
| *** jlhinson has joined #openstack-swift | 14:36 | |
| jrichli | mahatic: ok. that last patch in the chain is just adding account and container encryption, so it won't show you everything that has been added | 14:36 |
| mahatic | jrichli, okay. I haven't downloaded any code related to encryption (that hasn't been merged). Am I to start with this patch? | 14:37 |
| jrichli | mahatic - yes, start with that one. When you download it, the dependent patches will also be downloaded so you will see it all together. | 14:38 |
| mahatic | oh okay. sure | 14:38 |
| jrichli | mahatic - mainly focus on encrypter, decrypter, trivialkeymaster and their unit tests. those are the significant items that have been added. | 14:39 |
| mahatic | okay | 14:40 |
| *** lpabon has joined #openstack-swift | 14:47 | |
| *** hrou has quit IRC | 14:53 | |
| *** hrou has joined #openstack-swift | 14:54 | |
| notmyname | good morning | 14:54 |
| mahatic | good morning | 14:57 |
| wbhuber | good morning | 14:57 |
| notmyname | it's fun reading all the overnight (to me) conversations | 14:58 |
| notmyname | :-) | 14:58 |
| notmyname | clayg: thanks for fixing my pep8 | 14:59 |
| onovy | notmyname, hi, https://review.openstack.org/#/c/192562/4/AUTHORS should i remove my name, as paul suggests? thanks | 14:59 |
| *** haomaiwang has quit IRC | 15:00 | |
| notmyname | onovy: have you contributed to swift at all before, perhaps with a different email address? | 15:00 |
| onovy | no | 15:00 |
| notmyname | onovy: so, first off, thanks for the patch :-) | 15:00 |
| *** marzif has quit IRC | 15:00 | |
| onovy | your are welcome :) | 15:00 |
| notmyname | onovy: the AUTHORS update is fine | 15:00 |
| onovy | ok, thanks | 15:00 |
| *** haomaiwang has joined #openstack-swift | 15:01 | |
| notmyname | onovy: normally I do it all at once right before a release like peluse said, but it doesn't really matter when it happens | 15:01 |
| *** _hrou_ has joined #openstack-swift | 15:01 | |
| notmyname | and it seems you've done it "right" (in the right order instead of appended as some people do, and there's no reason to update the .mailmap file) so it makes my job easier | 15:02 |
| notmyname | so it's great | 15:02 |
| *** joeljwright has quit IRC | 15:02 | |
| notmyname | onovy: that's an interesting feature. I like it | 15:03 |
| notmyname | onovy: how are you using swift today? | 15:03 |
| *** aix has quit IRC | 15:03 | |
| *** maestro has quit IRC | 15:04 | |
| *** hrou has quit IRC | 15:04 | |
| *** hrou has joined #openstack-swift | 15:04 | |
| *** _hrou_ has quit IRC | 15:06 | |
| *** fthiagogv has joined #openstack-swift | 15:06 | |
| *** fthiagogv_ has joined #openstack-swift | 15:07 | |
| *** david-lyle has joined #openstack-swift | 15:09 | |
| notmyname | onovy: ok, after looking at it more, I'm slightly less excited about it | 15:09 |
| *** tsg has quit IRC | 15:11 | |
| notmyname | I'll leave comments in gerrit | 15:11 |
| *** proteusguy has quit IRC | 15:13 | |
| *** annegentle has quit IRC | 15:13 | |
| *** annegentle has joined #openstack-swift | 15:14 | |
| *** acorwin has quit IRC | 15:15 | |
| *** alpha_ori has quit IRC | 15:15 | |
| *** haomaiwang has quit IRC | 15:18 | |
| *** haomaiwang has joined #openstack-swift | 15:19 | |
| *** alpha_ori has joined #openstack-swift | 15:19 | |
| *** silor1 has joined #openstack-swift | 15:19 | |
| *** annegentle has quit IRC | 15:20 | |
| *** breitz has quit IRC | 15:21 | |
| *** blmartin has quit IRC | 15:21 | |
| *** breitz has joined #openstack-swift | 15:22 | |
| *** silor has quit IRC | 15:22 | |
| *** silor1 is now known as silor | 15:22 | |
| *** acorwin has joined #openstack-swift | 15:22 | |
| openstackgerrit | Merged openstack/swift: Fix Associated Projects link on Getting Started https://review.openstack.org/204167 | 15:24 |
| *** proteusguy has joined #openstack-swift | 15:25 | |
| *** silor has quit IRC | 15:31 | |
| *** zaitcev has joined #openstack-swift | 15:35 | |
| *** ChanServ sets mode: +v zaitcev | 15:35 | |
| *** chmouel has quit IRC | 15:44 | |
| *** aix has joined #openstack-swift | 15:46 | |
| *** aix has quit IRC | 15:46 | |
| *** Guest23745 has joined #openstack-swift | 15:47 | |
| *** nadeem has joined #openstack-swift | 15:47 | |
| *** Guest23745 has quit IRC | 15:47 | |
| *** nadeem has quit IRC | 15:48 | |
| *** nadeem has joined #openstack-swift | 15:49 | |
| *** tsg has joined #openstack-swift | 15:49 | |
| *** chmouel has joined #openstack-swift | 15:51 | |
| *** DericHorn-HP has quit IRC | 15:51 | |
| notmyname | onovy: review left | 15:51 |
| *** agentle_ has joined #openstack-swift | 15:52 | |
| clayg | notmyname: you can still be excited about a good idea that just needs a bit more work! | 15:53 |
| notmyname | well, I feel very conflicted about that | 15:53 |
| clayg | oic, you're only *slightly less* - so *still excited* | 15:53 |
| clayg | heh - about the feature? I haven't really looked at it - torgomatic commented "this is a good idea" - so I assume it's a great idea. | 15:54 |
| notmyname | yeah, that was my first reaction too | 15:54 |
| notmyname | I'm not forgeting some timestamp stuff we generate on the storage nodes, am I? | 15:55 |
| *** aix has joined #openstack-swift | 15:55 | |
| notmyname | I guess the reconciler does that now, right? but otherwise all timestamps are generated on the proxy, right? | 15:55 |
| notmyname | it's a great patch. small scope, docs (including manpages), tested functionality. I'm really happy about that | 15:57 |
| minwoob | clayg: Quick question about https://bugs.launchpad.net/swift/+bug/1393592 | 16:02 |
| openstack | Launchpad bug 1393592 in OpenStack Object Storage (swift) "proxy doesn't error limit nodes on put" [Undecided,New] | 16:02 |
| redbo | yeah, but object servers need to be fairly close. like for rejecting GETs to expired objects and removing tombstones | 16:02 |
| *** agentle_ has quit IRC | 16:03 | |
| notmyname | redbo: ah, yeah. good point | 16:03 |
| minwoob | clayg: It looks like the change needs to be made to swift.proxy.controllers.obj -- not sure where you're getting that the GET responses are incrementing the errors properly. | 16:04 |
| minwoob | clayg: Perhaps I've been looking at the wrong module? | 16:04 |
| *** rledisez has quit IRC | 16:04 | |
| minwoob | clayg: (instead of swift.proxy.controllers.obj). | 16:05 |
| *** blmartin has joined #openstack-swift | 16:09 | |
| *** gyee has joined #openstack-swift | 16:12 | |
| *** jistr has quit IRC | 16:19 | |
| *** breitz has quit IRC | 16:21 | |
| *** mahatic has quit IRC | 16:23 | |
| peluse | clayg, thanks for the tests.... about to roll those in. Refactored the ugly function quite a bit, am planning on moving the helpers out to module level as well to facilitate easier test coverage. Will let you know when ready for another look | 16:25 |
| *** breitz has joined #openstack-swift | 16:28 | |
| *** jordanP has quit IRC | 16:30 | |
| *** jlhinson has quit IRC | 16:33 | |
| *** DericHorn-HP has joined #openstack-swift | 16:34 | |
| *** DericHorn-HP has quit IRC | 16:35 | |
| *** ozone227 has joined #openstack-swift | 16:48 | |
| notmyname | meeting agenda for today is at https://wiki.openstack.org/wiki/Meetings/Swift. please update if you have other stuff to highlight, and check out is you have anything to report on | 16:48 |
| *** ozone227 has quit IRC | 16:48 | |
| *** jasondotstar has quit IRC | 16:50 | |
| *** silor has joined #openstack-swift | 16:53 | |
| *** silor has quit IRC | 16:57 | |
| *** silor has joined #openstack-swift | 16:59 | |
| clayg | minwoob: idk, I thought there was a patch one time where I did something about making error limiting better - maybe that bug is invalid? | 17:01 |
| clayg | peluse: you're welcome - hope they're helpful - please squash them as I'll probably write more tests as I review | 17:01 |
| clayg | peluse: i saw the refactor and I think that's on the right track - all that function(state) stuff did make me thinking of method(self) :D | 17:02 |
| *** wbhuber has quit IRC | 17:03 | |
| *** blmartin has quit IRC | 17:04 | |
| clayg | minwoob: 99d5018 maybe? | 17:04 |
| *** tsg has quit IRC | 17:04 | |
| *** silor1 has joined #openstack-swift | 17:07 | |
| *** wbhuber has joined #openstack-swift | 17:07 | |
| *** silor has quit IRC | 17:10 | |
| *** silor1 is now known as silor | 17:10 | |
| *** hseipp has left #openstack-swift | 17:12 | |
| jith_ | Hi all, i have configured the swift (SAIO) for devstack and used the seperate swift for glance backend.. but in horizon->system information , all the services are getting listed but the ip address and Enabled status of swift is not there.. how to make enable in horizon? | 17:12 |
| *** jasondotstar has joined #openstack-swift | 17:13 | |
| *** janonymous has joined #openstack-swift | 17:14 | |
| *** marzif has joined #openstack-swift | 17:15 | |
| *** jlhinson has joined #openstack-swift | 17:17 | |
| *** aix has quit IRC | 17:20 | |
| *** marzif has quit IRC | 17:25 | |
| *** marzif has joined #openstack-swift | 17:26 | |
| *** thurloat_isgone is now known as thurloat | 17:31 | |
| *** marzif has quit IRC | 17:31 | |
| *** marzif has joined #openstack-swift | 17:32 | |
| *** tab____ has joined #openstack-swift | 17:48 | |
| *** acorwin has quit IRC | 17:49 | |
| *** acorwin has joined #openstack-swift | 17:52 | |
| peluse | clayg, you there? | 17:53 |
| clayg | peluse: sure | 17:53 |
| peluse | so on the test you added test_ec_get_disk_layout()... the reason for the various # of responses is the basic logic of how I go after 'more'. Basically 2 rules | 17:54 |
| peluse | 1) if any of the initial ndata come back with 404, the regular node_iter hunts down more | 17:54 |
| peluse | 2) if they are all OK but I don't have enough etag sets, I go after parity nodes (all of them) | 17:54 |
| peluse | so your test has the first 10 all working OK (2xx) but one mismatched etag so I go after parity nodes, not hndoff since I don't suspect a handoff is oinvolved because there was no 404 | 17:55 |
| peluse | make sense sort of? | 17:55 |
| peluse | oh, 3rd rule. if there are alternates reporte, I go after those if I didn't get enough good responses.... | 17:56 |
| *** jasondotstar has quit IRC | 17:57 | |
| clayg | peluse: so... update the test? Or... update the code? | 18:01 |
| *** marzif has quit IRC | 18:02 | |
| *** mahatic has joined #openstack-swift | 18:02 | |
| peluse | clayg, I would say test since getting 210 2xx codes back but not having a set should not force us to the handoffs, there's no reason to beleive we'd find something there but we can be certain by issuing nparity reqeusts we'll likely get another FI that can be used to rebuild | 18:03 |
| peluse | ugh, not 210, just 10 | 18:03 |
| *** Kennan has joined #openstack-swift | 18:04 | |
| *** Kennan2 has quit IRC | 18:04 | |
| peluse | as a code change though, I could not issue all nparity requests, just the number that I need? | 18:04 |
| *** marzif has joined #openstack-swift | 18:04 | |
| peluse | guess I should draw a flow chart to help everyone see which nodes we check when < ndata and why.... | 18:05 |
| clayg | peluse: yeah I mean we could just issue requests to 2 x replica count nodes and then thow out stuff we don't need - but historically we've sort of tried to connect to just enough nodes to service the request - I think we tried roughly to carry that idea into EC? | 18:05 |
| clayg | peluse: I sort of imagine it'd be like while not have_enough_nodes(): connect_to_another_node() | 18:05 |
| peluse | clayg, yeah, right now if I need just 1 more I always go after nparity. Easy enough to change | 18:06 |
| clayg | peluse: oh... ok - i'm still digesting the code - so I'm trying to understand by poking at its behavior with tests | 18:06 |
| peluse | I'll play with is a bit and see how it shakes out... will let ya know | 18:06 |
| peluse | and I sooooo appreciate it too! | 18:06 |
| clayg | peluse: yeah sure! so but if you could squash in those tests (and you know change them so they work or are better or w/e) | 18:07 |
| peluse | I'l push a small update here soon, just moving the helpers to module level and probably this one tweak we just talked about | 18:07 |
| clayg | peluse: I'd just really like to have those test rigs in there as I review the change | 18:07 |
| peluse | OK, so how do I do the actual squash? Right now i just copied and pasted into my patch | 18:07 |
| *** marzif_ has joined #openstack-swift | 18:07 | |
| peluse | or is that OK? | 18:07 |
| clayg | peluse: yeah that wfm | 18:08 |
| peluse | coolio | 18:08 |
| peluse | they're in there, will be in next push here shortly.... | 18:08 |
| clayg | next time you push as long as when I git review -d I have all that StubResponse get_response with crazy_faked_out_logged_connections stuff I'm happy | 18:08 |
| clayg | peluse: cool no rush - I think I have notmyname's async container updaters on my review list today | 18:08 |
| notmyname | yay :-) | 18:09 |
| clayg | maybe some of kota's stuff from last night | 18:09 |
| clayg | OMG - EFFING versioned writes - need that so much | 18:09 |
| clayg | oh... and the meeting | 18:09 |
| peluse | yay, meetings | 18:09 |
| clayg | peluse: well it's a swift meeting | 18:10 |
| clayg | peluse: so it's roughly "hang out with my favorite peeps online and give notmyname a hard time" - which is kind of the best meeting | 18:10 |
| notmyname | also, it's kinda what you do every day ;-) | 18:10 |
| *** marzif has quit IRC | 18:11 | |
| *** marzif_ has quit IRC | 18:19 | |
| *** mahatic has quit IRC | 18:20 | |
| *** mahatic has joined #openstack-swift | 18:23 | |
| *** wbhuber has quit IRC | 18:25 | |
| *** agentle_ has joined #openstack-swift | 18:26 | |
| peluse | clayg, heh... so the change to go after just the right number of parity nodes (how many short we are of an etag set) works like a champ w/o having to change the test code but I'll annotate the code to show what its checking for so its clearer | 18:26 |
| *** silor1 has joined #openstack-swift | 18:34 | |
| *** wbhuber has joined #openstack-swift | 18:36 | |
| *** jasondotstar has joined #openstack-swift | 18:36 | |
| clayg | zaitcev: did you intend to merge commented asserts in https://review.openstack.org/#q,Ibab5c7605860ab768c8aa5a3161a705705689b04,n,z <- cschwede is keen to remove them in patch 204575 | 18:37 |
| patchbot | clayg: https://review.openstack.org/#/c/204575/ | 18:37 |
| *** silor has quit IRC | 18:38 | |
| *** silor1 is now known as silor | 18:38 | |
| zaitcev | clayg: I need to look, give me a moment | 18:38 |
| clayg | zaitcev: I was about to approve, but IIRC I've been known to leave commented stuff in tests because it makes it easier to debug the tests later - normally with a comment like "# this is useful for debugging" bug probably not always | 18:38 |
| mahatic | jrichli, hey, sorry I had stepped out for dinner. I did download the code. Could you tell me a little more on the probe test? | 18:39 |
| mahatic | jrichli, "the purpose is to have an integrated test with the real encryption/decryption middleware and the crypto class." You mean the current unit test covers only the encrypter middleware and does not test crypto? | 18:40 |
| clayg | mahatic: are you working on probetests for encryption!? | 18:40 |
| mahatic | clayg, :D that's the plan ;) I started looking at it. Any help would be great! | 18:41 |
| jrichli | there are some tests for each piece individually : so for example, the unit tests against the encrypter only test the encrypter, and mocks the other pieces | 18:41 |
| clayg | mahatic: did you by any chance setup encryption using the crypto branch -> https://github.com/swiftstack/vagrant-swift-all-in-one/tree/crypto | 18:41 |
| clayg | it does all the footer middleware stuff for you and the like | 18:41 |
| clayg | it'd be nice to have any crypto specific updates merged there to help other reviews of crypto come on board more quikcly | 18:41 |
| mahatic | jrichli, right. but the probe test is expected to test it all together? | 18:42 |
| clayg | jrichli: didn't we have a prioirtized list of big ticket outstanding unresolved issues for encyrption? | 18:42 |
| clayg | jrichli: I didn't think that probetests was on the top of the list? | 18:42 |
| jrichli | clayg: I didn't know that vagrant was out there. thanks! I can look into pointing it to the latest. | 18:42 |
| mahatic | clayg, nope. I just downloaded the code from here: https://review.openstack.org/#/c/203454/ | 18:43 |
| jrichli | clayg: There is a list of outstanding issues. The problem is, there is design that has to happen in the high priority issues. | 18:43 |
| clayg | mahatic: right so the code doesn't work unless you do the configuration stuff - the crypto branch can help you - or i'm sure jrichli has written down somewhere how to do it by hand | 18:43 |
| clayg | jrichli: stupid design | 18:44 |
| clayg | jrichli: where's acoles when you need him | 18:44 |
| clayg | @channel - does anyone in here know how to design software!? | 18:44 |
| jrichli | :-) Yes, I think we can get more movement on some of those items when he returns. | 18:45 |
| mahatic | :) | 18:45 |
| jrichli | lol | 18:45 |
| jrichli | clayg: I am starting to take a stab at real footer support, but it will be challenging for me | 18:45 |
| clayg | jrichli: you got this ;) | 18:45 |
| jrichli | thx! | 18:46 |
| zaitcev | clayg: commented on patch 204575 | 18:47 |
| patchbot | zaitcev: https://review.openstack.org/#/c/204575/ | 18:47 |
| jrichli | mahatic: It's probably best to start on the config and get encryption working on you SAIO. We will make sure you get the same results when running functsts | 18:47 |
| clayg | zaitcev: oh lol - they weren't even your commented assertions | 18:48 |
| clayg | zaitcev: goodness - I just pass the buck to the first core in the blame :P | 18:48 |
| zaitcev | clayg: no, they were | 18:48 |
| zaitcev | clayg: but probably bitrotted by now | 18:48 |
| clayg | zaitcev: ok cool - so let's pull 'em! | 18:49 |
| clayg | zaitcev: thanks! | 18:49 |
| mahatic | jrichli, sure. Now that I downloaded the code, any doc or something that points to config stuff? | 18:49 |
| clayg | mahatic: *ahem* crypto branch of vagrant-swift-all-in-one </shameful-plug> | 18:49 |
| mahatic | jrichli, or clone it from the vagrant one? | 18:49 |
| *** silor has quit IRC | 18:50 | |
| jrichli | mahatic: look at etc/proxy-server.conf-sample - and follow the instructions for adding the new middleware | 18:50 |
| mahatic | clayg, lol. you also pointed that jrichli would have something, I was just checking | 18:51 |
| jrichli | plus, add "fake_footers" middleware with "use = egg:swift#fake_footers" | 18:51 |
| jrichli | so the additions to the pipeline are: | 18:52 |
| jrichli | decrypter trivialkeymaster encrypter fake_footers - and put those between dlo and proxy-logging | 18:52 |
| clayg | jrichli: when you lay it all out like that it seems silly to have a whole branch *just* for that :\ | 18:54 |
| jrichli | clayg: maybe there will be more later :-) | 18:54 |
| clayg | jrichli: that's the ticket! | 18:55 |
| mahatic | jrichli, can I have your config file? a paste? | 18:56 |
| jrichli | ok, one min | 18:56 |
| openstackgerrit | paul luse proposed openstack/swift: Add ability for GET path to see/select alternate frag archs https://review.openstack.org/201283 | 18:58 |
| jrichli | mahatic: this version sets DEBUG level logging for the new middleware http://paste.openstack.org/show/400888/ | 19:00 |
| *** jlhinson has quit IRC | 19:01 | |
| *** jlhinson has joined #openstack-swift | 19:01 | |
| *** mahatic_ has joined #openstack-swift | 19:12 | |
| *** changbl has joined #openstack-swift | 19:12 | |
| *** mahatic has quit IRC | 19:13 | |
| *** mahatic has joined #openstack-swift | 19:14 | |
| clayg | can anyone else try to help out david on lp bug #1477229 | 19:17 |
| openstack | Launchpad bug 1477229 in OpenStack Object Storage (swift) "Adding proxy-logging to the pipeline is not enough for access logging" [Undecided,New] https://launchpad.net/bugs/1477229 | 19:17 |
| clayg | I feel like I keep saying - no dude what you're saying makes no sense; this works for everyone; need more info - and he's like - yeah it's totally broken, so do whatever you want | 19:18 |
| *** mahatic_ has joined #openstack-swift | 19:19 | |
| *** mahatic has quit IRC | 19:20 | |
| mahatic_ | jrichli, okay done | 19:21 |
| jrichli | try running the functests and see how many errors and failures you get | 19:21 |
| mahatic_ | sure | 19:22 |
| mahatic_ | jrichli, Entry point 'decrypter' not found in egg 'swift' - what i get when do a startmain. I do have a an entry in proxy-server.conf | 19:27 |
| mahatic_ | am I missing something? | 19:27 |
| jrichli | did you run 'sudo python setup.py develop' and 'sudo python setup.py install' before trying to run? | 19:28 |
| mahatic_ | jrichli, some issue there. Just noticed | 19:30 |
| mahatic_ | proxy stuff | 19:30 |
| jrichli | what is the error? | 19:32 |
| mahatic_ | no, this proxy is internal, intel. I'm running them again | 19:33 |
| *** jasondotstar has quit IRC | 19:36 | |
| jrichli | clayg, mahatic_ - speaking of internal, eventually you will also add this new middleware to the internal client pipeline. But you can do that later. | 19:37 |
| clayg | jrichli: heh - yeah good point - i mean it depends on the keymaster right? | 19:38 |
| clayg | jrichli: like some keymaster may require a user token to get a key right? like public access to encrypted objects won't work? | 19:38 |
| clayg | jrichli: or are we not going to have keymasters like that? | 19:38 |
| jrichli | clayg: I still dont know much about how the internal client is used (other than it making internal requests in general) | 19:40 |
| glange | we use it for log processing | 19:40 |
| jrichli | clayg: as for the keymaster needing a token, that depends. the keystone token can be used | 19:41 |
| jrichli | glange: ok, thanks. Is it also used for other things? | 19:42 |
| glange | not sure, maybe some little tools that people have written, but the only big thing is log processing | 19:44 |
| notmyname | no internal client usage in container sync? | 19:44 |
| hrou | storlets use it ; ) | 19:45 |
| glange | yeah, looks like it is in there, but I assumed jrichli knew about all the swift codebase uses :) | 19:45 |
| glange | just grep! | 19:45 |
| jrichli | glange: bad assumption :-) | 19:46 |
| glange | haha | 19:46 |
| glange | step 1, grep swift codebase! | 19:46 |
| notmyname | leaping to conclusions is more fun | 19:46 |
| jrichli | glange: yes. i haven't been focused on digging into the internal client use yet | 19:46 |
| jrichli | glange: just figured i'd ask while we were on the topic | 19:47 |
| glange | ok | 19:47 |
| glange | we use it sort of like notmyname's internal_proxy in https://github.com/notmyname/slogging | 19:47 |
| jrichli | and, now I know storlets use it - which will mean it needs the middleware | 19:48 |
| notmyname | glange: which was the exact same thing at one point ;-) | 19:48 |
| *** fthiagogv_ has quit IRC | 19:58 | |
| *** marzif_ has joined #openstack-swift | 20:05 | |
| *** jasondotstar has joined #openstack-swift | 20:12 | |
| clayg | since when does replication pick up a .data file delction without a hash invalidation? | 20:20 |
| clayg | the every-so-often do_listdir is for the suffix dirs - to make sure the keys in the hash match the dirs in the part | 20:21 |
| openstackgerrit | Hamdi Roumani proposed openstack/swift: Updated tox setup instructions on Development Guidelines https://review.openstack.org/204749 | 20:26 |
| mahatic_ | jrichli, functests -> FAILED (SKIP=12, errors=78, failures=34) ^ | 20:29 |
| jrichli | that is more than what I get, so something is wrong | 20:30 |
| openstackgerrit | Carlos Cavanna proposed openstack/swift: Improving statistics sent to Graphite. https://review.openstack.org/202657 | 20:30 |
| mahatic_ | hmm, what's expected? | 20:31 |
| jrichli | mahatic_: I will give you a link to a paste with the list of expected problems. you can look into some that aren't on that list to find out what's going on | 20:31 |
| mahatic_ | sure | 20:31 |
| jrichli | mahatic_ : http://paste.openstack.org/show/401160/ has the expected errors/failures | 20:35 |
| mahatic_ | jrichli, looking, thank you | 20:37 |
| mahatic_ | jrichli, off to bed, will dig in tomm. thanks for the pointers! | 20:37 |
| *** jasondotstar has quit IRC | 20:38 | |
| jrichli | mahatic_: ok. np! good night | 20:38 |
| *** fthiagogv has quit IRC | 20:40 | |
| *** mahatic_ has quit IRC | 20:43 | |
| *** jeblair is now known as CarolChanning | 20:45 | |
| *** CarolChanning is now known as jeblair | 20:45 | |
| *** ho has joined #openstack-swift | 20:48 | |
| ho | good morning! | 20:50 |
| mattoliverau | Morning | 20:52 |
| clayg | we doing this thing!? | 20:53 |
| *** hrou has quit IRC | 20:53 | |
| ho | mattoliverau: morning! | 20:53 |
| janonymous | o/ | 20:54 |
| clayg | i know that eventlet greenthreads aren't racy like os threads wrt to statements like "a += 1" because of "cooperative" yielding and all that | 20:55 |
| clayg | but... pyhton has a thing with syncronization that will let you do atomic increments right? | 20:55 |
| clayg | I thought it was collections.Counter - but that's totally a different thing | 20:55 |
| redbo | w/in 3 | 20:56 |
| clayg | ohhh.. maybe itertools.count() | 20:56 |
| *** agentle_ has quit IRC | 20:57 | |
| *** tongli has quit IRC | 20:57 | |
| *** tongli has joined #openstack-swift | 20:57 | |
| *** blmartin has joined #openstack-swift | 20:58 | |
| clayg | gd, if zzzeek doesn't know https://twitter.com/zzzeek/status/99899671262150656 | 20:58 |
| *** kota_ has joined #openstack-swift | 20:58 | |
| *** ChanServ sets mode: +v kota_ | 20:58 | |
| *** blmartin has quit IRC | 20:59 | |
| kota_ | morning | 20:59 |
| *** blmartin has joined #openstack-swift | 20:59 | |
| mattoliverau | kota_: morning | 20:59 |
| ho | kota_: morning! | 21:00 |
| notmyname | hello. last meeting just ended | 21:00 |
| notmyname | time for swift meeting in #openstack-meeting | 21:00 |
| *** DericHorn-HP has joined #openstack-swift | 21:03 | |
| *** agentle_ has joined #openstack-swift | 21:08 | |
| clayg | oh man I don't think it is safe - and weirdly - it goes to fast?! https://gist.github.com/clayg/d3630eeb3b6b680d631a | 21:08 |
| *** tab____ has quit IRC | 21:12 | |
| *** david-lyle has quit IRC | 21:17 | |
| *** marzif_ has quit IRC | 21:25 | |
| *** geaaru has quit IRC | 21:27 | |
| *** hrou has joined #openstack-swift | 21:30 | |
| *** tongli has quit IRC | 21:37 | |
| *** tongli has joined #openstack-swift | 21:37 | |
| *** agentle_ has quit IRC | 21:43 | |
| *** tongli has quit IRC | 21:46 | |
| *** annegentle has joined #openstack-swift | 21:46 | |
| *** tongli has joined #openstack-swift | 21:46 | |
| *** jkugel has quit IRC | 21:46 | |
| *** tongli has quit IRC | 21:48 | |
| *** thurloat is now known as thurloat_isgone | 21:52 | |
| *** Nyny has joined #openstack-swift | 21:58 | |
| Nyny | Hello, it's possible to hide XML list file? | 21:58 |
| *** aerwin has joined #openstack-swift | 22:00 | |
| notmyname | Nyny: not sure I understand that question | 22:01 |
| mattoliverau | k, breakfast time. bbl | 22:01 |
| tdasilva | i'm guessing that's an apple swift question | 22:01 |
| *** jasondotstar has joined #openstack-swift | 22:01 | |
| notmyname | oh! maybe | 22:01 |
| notmyname | Nyny: are you talking about the iOS programming language? | 22:01 |
| MooingLemur | unless container listing in format=xml | 22:01 |
| notmyname | Nyny: we work on openstack-swift. the object storage system | 22:01 |
| *** Kennan has quit IRC | 22:02 | |
| Nyny | yes i speak about that, i have object storage on a provider and i would like to hide XML file that list all file in my container | 22:02 |
| tdasilva | sounds like austin hackathon agenda is going to be packed | 22:02 |
| ho | clayg: yesterday, I got your comment for patch 202411. I would like to have more comments which parts are wrong. I think you pointed out is not only the file name but alsomatrix testing??? | 22:03 |
| patchbot | ho: https://review.openstack.org/#/c/202411/ | 22:03 |
| *** Kennan has joined #openstack-swift | 22:03 | |
| clayg | ho: maybe the matrix just needed some comments - which keys are what | 22:03 |
| clayg | ho: yeah giventhe option of a file named test_account_access_control.py and test_axc.py - I always prefer the one that means something. | 22:04 |
| *** blmartin has quit IRC | 22:05 | |
| ho | clayg: There are some comment on the patch L21-L47. kind of this? | 22:05 |
| Nyny | notmyname : hide this : http://goopics.net/data/532o87.png | 22:05 |
| *** jrichli has quit IRC | 22:05 | |
| ho | clayg: And L21-51 is correct one. | 22:06 |
| clayg | ho: oic - yeah just move that closer to the definition | 22:06 |
| MooingLemur | Nyny: depending on the X-Container-Read setting, I suspect you just need to remove .rlistings | 22:06 |
| clayg | ho: another option is use dictionaries - since they're self documenting | 22:07 |
| clayg | ho: either way it's an amazing matric - so many scenarios! | 22:08 |
| *** aerwin has quit IRC | 22:09 | |
| ho | clayg: I used dic for my first version but a line become too long so i decided to use tupple and list then convert it to dic for a line. | 22:09 |
| clayg | ho: but it looks like some of the values are always None? | 22:09 |
| Nyny | ok thanks im going to look for | 22:09 |
| ho | clayg: yeah, i thought it might use in the future. i have no plan to use "account_name" so far because it's decided by keystone. | 22:10 |
| clayg | ho: yeah idk, that's the thing that DAMP is talking about - DRY code is not about saving lines, it's about maintainability - I feel like a dictionary where un-used keys can be omitted would be easier to read/understand - and therefore - easier to maintain - but it's just a though | 22:10 |
| ho | clayg: i see. today i will finish to write the other functional test for container acl after that if account_name is not used, i will remove it. | 22:12 |
| clayg | ho: up to you - whatever feels like it will be easist to review/maintain - use your best judgement - just thowing out something to think about - thanks for working on this | 22:13 |
| ho | clayg: thanks for the comments. I will think about it :-) | 22:14 |
| clayg | I've had this thing happen to me - and antadotically other software engineers have confessed it's a shared experience - where you start with two things and recognize they can be generalized to a common base with some paramaters or subclasses - whatever ... | 22:15 |
| clayg | then you have more things - eventually maybe half a dozen things thare are all basically variations of this common base that keeps getting smarter and smarter | 22:16 |
| clayg | ... then finally you add the next thing and it starts getting complicated - eventually to the point where your generalized thing is special casing one off options just to support one of it's dependents | 22:16 |
| *** wbhuber has quit IRC | 22:16 | |
| clayg | ... eventaully you have to break it up - pull it back apart into seperate things that you can maintain individually | 22:17 |
| clayg | (we sort went though this with some of the DiskFile and ObjectController stuff in swift) | 22:17 |
| clayg | taking the duplicate code and making it DRY has a good name - but taking the generalized code and pulling it apart (to my knowledge) - doesn't have a name? | 22:18 |
| *** jasondotstar has quit IRC | 22:18 | |
| *** jlhinson has quit IRC | 22:18 | |
| Nyny | MooingLemur: remove .rlistings from the container? | 22:19 |
| clayg | maybe someon who's read gang of four more througly can remind me of a concept i'm forgetting? | 22:19 |
| Nyny | it's a file? | 22:19 |
| clayg | Nyny: it's an option to X-Container-Read-Acl | 22:20 |
| notmyname | Nyny: do you want to get rid of one item or are you wanting to turn off the listing entirely? | 22:20 |
| clayg | Nyny: do you have the container set to be publicly accessible? ".r:*" | 22:21 |
| Nyny | just turn off the listing | 22:21 |
| dfg_ | clayg: i found my saio problem. it was that my python-six was too old. and pip wouldn't upgrade it because it had been isntalled by the system or something. i had to apt-get remove python-six and then pip install it and it worked. | 22:21 |
| Nyny | yes i have a public container accessible by anyone | 22:21 |
| clayg | dfg_: YOU'RE AWESOME - thanks for the report! | 22:22 |
| ho | clayg: yeah, I'm in front of complexity. gang of four sounds like peple who made 20? design patterns | 22:22 |
| clayg | dfg_: does that mean you're up and running on *precise* !? | 22:22 |
| clayg | ho: yeah that's the stuff! | 22:22 |
| dfg_ | clayg: ya i guess. you do feel less safe? | 22:22 |
| dfg_ | do you | 22:22 |
| clayg | dfg_: I live in constant terror that dfg_ might get swift working on precise | 22:23 |
| dfg_ | :) | 22:23 |
| clayg | lol | 22:23 |
| clayg | ok - so itertools.count - not thread safe - terrible | 22:24 |
| clayg | back to testing when async container updates cause observalbe incosistencies in the container listings | 22:24 |
| * notmyname lives in constant terror that clayg will review his code | 22:25 | |
| ho | clayg: maybe "Patterns and Antipatterns" (Andrew Koenig) | 22:25 |
| notmyname | (but it's better for it) | 22:25 |
| clayg | notmyname: gholt was the worst - you could do something that is by all accounts pretty decent working testing and documented code | 22:26 |
| *** jasondotstar has joined #openstack-swift | 22:27 | |
| clayg | not only would he find half a dozen nits that you are terribly embarassed for submitting - but bugs - over and over again he'd point out my BUGS | 22:27 |
| clayg | :'( | 22:27 |
| clayg | definately better for it | 22:27 |
| notmyname | yes. it happened to me too :-( | 22:27 |
| * clayg misses gholt | 22:28 | |
| clayg | omg is swift client not green :'( | 22:33 |
| clayg | guess I'll need that thread safe counter after all | 22:33 |
| *** jasondotstar has quit IRC | 22:33 | |
| *** minwoob has quit IRC | 22:36 | |
| notmyname | clayg: wehre are you seeing that? | 22:36 |
| Nyny | if X-Container-Read is not set, container listing access is public? | 22:37 |
| notmyname | http://logs.openstack.org/50/89250/111/check/gate-python-swiftclient-python26/0153d6f/testr_results.html.gz ? | 22:37 |
| clayg | notmyname: i have a script where I tried to make a swiftclient.Connction in a gt and it's not patched - i guess if i monkey patched it it would just work? | 22:38 |
| clayg | oh oh - green like eventlet green | 22:38 |
| notmyname | ah ok | 22:38 |
| clayg | notmyname: :p | 22:38 |
| clayg | Nyny: container listing access is NOT public | 22:38 |
| clayg | Nyny: unless you're not running auth? | 22:38 |
| clayg | Nyny: some auth systems' responses will include a Set-Cookie header - so if you're passing auth via your web browser I don't think there's a super great way to dissallow it | 22:39 |
| clayg | but if you don't have a read acl set and you get that xml in the browser you should verify real quick that visiting that url in a private browser session will 4XX | 22:40 |
| Nyny | my provider run with auth | 22:41 |
| Nyny | i try with private browser session, file listing still public | 22:42 |
| clayg | sounds like acl or no auth to me | 22:42 |
| clayg | Nyny: can you check other containers? | 22:42 |
| clayg | Nyny: do you *want* the objects to public and just not the listing? how did you accomplish this if not acls - most auth systems limit all requets to private authenticated users - so what's going on? | 22:43 |
| Nyny | public objects (when i create container on my provider panel, i select "public container"), hide listing | 22:44 |
| clayg | hrm.... | 22:45 |
| *** DericHorn-HP has quit IRC | 22:45 | |
| clayg | Nyny: i don't get it - start validating things with curl | 22:47 |
| clayg | Nyny: when the web doesn't make sense - you've always got curl | 22:47 |
| clayg | long live curl | 22:47 |
| clayg | zackmdavis: I don't want to hear about httpie | 22:48 |
| torgomatic | curl means never having to say "wait, what requests was that sending again?" | 22:48 |
| notmyname | torgomatic: except for unicode ;-) | 22:48 |
| notmyname | but thats the shell/terminal that's being "helpful" | 22:49 |
| torgomatic | wireshark means... | 22:49 |
| clayg | I get it that ascii wasn't going to cut it - but we could have stopped with utf-8 - just been like "nope, that's enough with the charaters" | 22:49 |
| zackmdavis | clayg, I've never used httpie; did you mean to tab-complete someone else's name? | 22:49 |
| clayg | zackmdavis: oh - sorry - i guess when I think of the kids new "interesting" hipster tools I think of zackmdavis | 22:50 |
| zackmdavis | clayg, buuuuut now that you mention it, this looks pretty cool https://github.com/jkbrzt/httpie; maybe I'll try it | 22:51 |
| clayg | LOLLO | 22:51 |
| clayg | i *just* said to myself "i get he'll go try it now" | 22:51 |
| clayg | s/get/bet/ | 22:51 |
| *** DericHorn-HP has joined #openstack-swift | 22:51 | |
| clayg | zackmdavis folks, a scolar and a commedian - it's like torgomatic but without the jolly santa claus vibe | 22:52 |
| *** DericHorn-HP has quit IRC | 22:53 | |
| torgomatic | my beard doesn't have *that* much gray in it | 22:53 |
| torgomatic | probably | 22:53 |
| clayg | notmyname: ok in a trival test default is golden and 0.01 is observable pretty quickly :) | 23:01 |
| clayg | this contaienr has ~20K rows in it | 23:02 |
| notmyname | yay | 23:02 |
| clayg | it's just a development environment - but the idea I think that we can go pretty low and be fine as long as we don't turn off the waiting is pretty solid | 23:02 |
| clayg | nice work! | 23:02 |
| ho | all, in the meeting jrichli said acoles is going to back next week..., is he in vacation now? | 23:04 |
| clayg | ho: not really - I think he's like doing his job or something? | 23:05 |
| openstackgerrit | Samuel Merritt proposed openstack/swift: functests: use assertIn and assertNotIn https://review.openstack.org/204785 | 23:06 |
| ho | clayg: thanks! now i understand the reason why i can not get in touch with hime :-) | 23:06 |
| clayg | torgomatic: man you're all about the death to python 2.6 | 23:06 |
| torgomatic | clayg: yeah, I do stuff like that when I get hung up on other problems. Feels like progress toward *something* at least. | 23:07 |
| Nyny | control list (ACL) is is that? http://goopics.net/i/532o87 | 23:07 |
| torgomatic | I didn't notice it until I reviewed janonymous's patch to use assertTrue instead of assertIn; a lot of those assertions were pretty stupid-looking if you didn't have to worry about 2.6 | 23:08 |
| clayg | notmyname: nice | 23:08 |
| clayg | Nyny: no it's in the headers - not the body | 23:08 |
| clayg | chrome and firefox both have nice tools to inspect the requests if you'd rather than that curl it might get you along the way | 23:09 |
| *** thurloat_isgone is now known as thurloat | 23:09 | |
| Nyny | set .rlistings in the header? | 23:10 |
| Nyny | to remove listing? | 23:11 |
| Nyny | booleen value? | 23:11 |
| clayg | Nyny: http://docs.openstack.org/developer/swift/misc.html#acls | 23:11 |
| clayg | X-Container-Read is the header name, and the value is a csv with a weird arcane syntax | 23:12 |
| *** km has joined #openstack-swift | 23:12 | |
| clayg | notmyname: 500ms looks like the magic number | 23:12 |
| clayg | notmyname: well.. that's pleanty anyways - what did you go with 100ms? | 23:13 |
| notmyname | ho: clayg: he was doing other HP stuff, but I think he's on vacation right now | 23:13 |
| clayg | notmyname: ah no 1000ms - perfect | 23:13 |
| clayg | notmyname: oh he is - nice work acoles_away ! | 23:14 |
| notmyname | clayg: what's magic about 500ms? | 23:14 |
| torgomatic | should be 1024ms; people won't question your made-up numbers if they're powers of 2 | 23:14 |
| notmyname | I chose 1s because, we'll it's probably in the goldilocks zone for the behavior | 23:14 |
| ho | clayg: +1 for acoles_away's vacation :) | 23:15 |
| clayg | and an underpowered swift all in one trying to find a inconsistent listing after a successful PUT - it's just hard to do with a 500ms container_update_timeout | 23:16 |
| clayg | notmyname: 1s is goldilocks for smoothing out the update timeouts when the container go to eat the .pending? | 23:16 |
| notmyname | ah, cool | 23:16 |
| clayg | notmyname: i forget what I did to test the container - oh yeah I had a timerouter_server.py | 23:17 |
| clayg | notmyname: let me try that again one more time, double check tests and put up the review - looks good | 23:17 |
| notmyname | 1s is a guess. it's shorter than 3 and "reasonable" for being good enough for functionling systmes | 23:17 |
| notmyname | mostly I chose it because I didn't want the bikeshed about what it should be ;-) | 23:18 |
| clayg | notmyname: ok - well I think your intuation was good | 23:18 |
| clayg | notmyname: well - let's see | 23:18 |
| *** jrichli has joined #openstack-swift | 23:19 | |
| Nyny | My X Container Header : X-Container-Read: .r:*,.rlistings, how curl request to remove rlistings? | 23:20 |
| clayg | curl -H 'x-auth-token: <your-token>' https://<your-storage-domain>/v1/<your-account>/<your-continer>/ -XPOST -H 'x-container-read: .r:*' should do it | 23:22 |
| clayg | is the trailing comma in the header? is that the last value? you have to write the whole acl you want as it will overwrite what's there | 23:23 |
| *** jasondotstar has joined #openstack-swift | 23:25 | |
| Nyny | it works thanks ! | 23:27 |
| clayg | Nyny: do you think that's something you could help your provider change the default behavior with? | 23:28 |
| clayg | Nyny: or as a user of swift is there somewhere you originally looked for documentation on how to use the acl api that could have included more inforatmion about this topic? | 23:29 |
| clayg | Nyny: neway - glad you got it sussed - happy to help | 23:29 |
| *** jlhinson has joined #openstack-swift | 23:31 | |
| Nyny | On my provider panel im allowed to create different profil container : private, public and static, anyone have a different comportment | 23:31 |
| Nyny | just public have rlistings by default | 23:32 |
| clayg | i wonder what static is? | 23:32 |
| clayg | notmyname: ^ do you have any idea which provider this is? | 23:32 |
| Nyny | for website, heberg static content | 23:32 |
| Nyny | css, img, js | 23:32 |
| Nyny | optimize for, i thinksx | 23:33 |
| clayg | Nyny: maybe that's public w/o listings? | 23:33 |
| Nyny | probably | 23:33 |
| *** jlhinson has quit IRC | 23:33 | |
| notmyname | nope | 23:33 |
| ho | clayg: sounds like hp's cloud. | 23:33 |
| notmyname | no, HP has "private" and "public" only | 23:34 |
| ho | clayg: they provide staticweb with CDN. | 23:34 |
| Nyny | my provider : http://www.ovh.com/fr/cloud/storage/object-storage.xml | 23:35 |
| ho | notmyname: i see. | 23:35 |
| notmyname | ah, ovh | 23:35 |
| clayg | Nyny: awesome! | 23:35 |
| Nyny | awesome? you like it? | 23:35 |
| ho | great! ovn has OpenStack logo. | 23:36 |
| notmyname | ovh actually has a pretty big swift deployment | 23:36 |
| clayg | Nyny: I like everyone who's using the open source storage engine built for the web - death to our amazon overloads | 23:36 |
| clayg | I say as my kids are watching prime in the other room and I'm tying into a irssi screen running on a ec2 instance | 23:36 |
| clayg | but w/e | 23:37 |
| notmyname | lol | 23:37 |
| *** kei_yama has joined #openstack-swift | 23:37 | |
| notmyname | runabove (a subsidiary of ovh) has the same thing: http://d.not.mn/runabove_container_type.png | 23:37 |
| clayg | mattoliverau: dfg_: you think I could get some of that free rackspace could action? | 23:38 |
| clayg | notmyname: so *is* static public w/o the listings? | 23:38 |
| notmyname | oh, wow. "static" is a lot more than that (haven't checked acls yet) | 23:38 |
| clayg | cause that would make a lot of sense acctually | 23:38 |
| clayg | oic - k nm | 23:38 |
| notmyname | it has a whole bunch of default css/html/png files that populate it by default | 23:39 |
| clayg | omg crazy - neat tho | 23:39 |
| clayg | i love it when people build cool stuff on swift! | 23:39 |
| notmyname | static.AUTH-85633419797e494a8b0184a846a9bd4f.storage.bhs-1.runabove.io | 23:39 |
| notmyname | that's the one I just created | 23:39 |
| notmyname | look familiar? ;-) | 23:39 |
| notmyname | importantly, check out what's in it | 23:39 |
| Nyny | +notmyname: what it it? | 23:40 |
| notmyname | but it doesn't have staticweb set up! | 23:40 |
| notmyname | http://static.auth-85633419797e494a8b0184a846a9bd4f.storage.bhs-1.runabove.io/index.html | 23:40 |
| notmyname | Nyny: that's a normal swift container listing | 23:40 |
| notmyname | the xml | 23:40 |
| notmyname | well, it's xml instead of plaintext or json | 23:40 |
| clayg | notmyname: browsers love to accept the xml | 23:41 |
| notmyname | oh yeha | 23:41 |
| notmyname | that's why | 23:41 |
| clayg | http://static.auth-85633419797e494a8b0184a846a9bd4f.storage.bhs-1.runabove.io/?format=json | 23:41 |
| clayg | this is FUN! | 23:41 |
| clayg | SWIFT IS SO COOL! | 23:41 |
| notmyname | yup | 23:41 |
| notmyname | lol | 23:41 |
| notmyname | this is why I sign up with all the swift hosters ;-) | 23:41 |
| notmyname | ok, let me dig my creds out and see what the acls are | 23:42 |
| clayg | notmyname: thanks for doing that - it's a hoot to see what people are building out of this little project we get to play with | 23:42 |
| notmyname | no kidding! :-) | 23:42 |
| Nyny | on ovh.com, my static container don't have a file listing | 23:43 |
| clayg | i really like the dns stuff - i wish I had a good idea how to productize that for people - it's a great integration feature | 23:43 |
| Nyny | and he directly redirect to the index like an heberg storage | 23:43 |
| clayg | notmyname: but like what are we gunna do - "yeah just give us creds to your bind servers and we'll write our you zone files - it'll be cool - trust me" | 23:43 |
| Nyny | exept if i failed x) | 23:44 |
| notmyname | here's that container from runabove https://gist.github.com/notmyname/5c272e72acfb06d10d21 | 23:45 |
| notmyname | cool. they're setting up staticweb, CORS, ACLs, and the default obejcts | 23:45 |
| Nyny | yes you've right, static content make an xml listing, my mistake | 23:45 |
| notmyname | why isn't staticweb working there? | 23:45 |
| Nyny | what's the use of ".r:*" in the header? | 23:46 |
| notmyname | Nyny: makes it public | 23:47 |
| *** jlhinson has joined #openstack-swift | 23:47 | |
| notmyname | the long-form is ".referer:*" which means that any referer header is allowed, ie public | 23:47 |
| Nyny | ha ok, ty | 23:47 |
| clayg | notmyname: idk, maybe the middlware isn't in the pipeline? seems like they're setting up all the headers? | 23:47 |
| notmyname | to the SAIO, batman! | 23:48 |
| mattoliverau | clayg: sorry been in a meeting and about to head into another, I'll try and find someone to ping about that ;) | 23:51 |
| notmyname | hmm...works for me on the SAIO | 23:52 |
| notmyname | gives me index.html | 23:52 |
| notmyname | 404 if index.html isn't found | 23:52 |
| Nyny | and what do you think about the OVH Object Storage price versus competitor (like amazon)? 0,01e per Go storage and 0,01e per Go output bandwidth | 23:52 |
| clayg | notmyname: i'd say it's not in the pipeline | 23:53 |
| clayg | notmyname: it generates extra requests sometimes - they might have pulled it | 23:53 |
| notmyname | clayg: it's in /info | 23:54 |
| ho | notmyname, clayg: http://gaul.org/object-store-comparison/ | 23:54 |
| clayg | notmyname: oh | 23:54 |
| *** dmorita has joined #openstack-swift | 23:55 | |
Generated by irclog2html.py 2.14.0 by Marius Gedminas - find it at mg.pov.lt!
