clayg | yeah | 00:01 |
---|---|---|
tdasilva | clayg: sorry, I see what you mean, will change the code | 00:04 |
*** ho has joined #openstack-swift | 00:04 | |
ho | good morning guys! | 00:08 |
clayg | peluse: acoles_away: I'm looking at 159637 - since it's not marked WIP | 00:13 |
peluse | clayg, cool. its ready for eyes | 00:15 |
peluse | the other one is a little bit of a mess due to changes needed because of the multi-phase put stuff | 00:15 |
clayg | and i'm gathering we're just not there on the etag stuff - tdasilva seems to have a patch against feature/ec that starts that work? | 00:17 |
peluse | clayg, container DB update using correct etag was the only thing I knew of | 00:18 |
clayg | peluse: the proxy also returns some random FA's etag on PUT | 00:19 |
tdasilva | clayg, peluse: i'm having some issues with my dev. environment with ec work, so i'm trying to get that resolved | 00:19 |
clayg | peluse: i found it easiest to see with a upload via curl using -T /path/to/local/file - the md5 will come back different every time | 00:20 |
peluse | clayg, yeah, well I meant the only thing I knew about until you mentioned the md5 test earlier | 00:22 |
peluse | but right now my whole dev env is jacked all to hell also so I'm ready to throw something | 00:22 |
clayg | peluse: oic, ok well those two things are the only ones I know about | 00:22 |
tdasilva | peluse: mmm...you wouldn't be running into issues with ec master, would you? | 00:24 |
peluse | no, my older VM had an old eventlet that didn't have the changes tsg made and the requirements file on feature/ec isn't up to snuff there yet | 00:26 |
peluse | running tests now.... | 00:26 |
tdasilva | ok | 00:26 |
*** lpabon has joined #openstack-swift | 00:28 | |
*** lpabon has quit IRC | 00:28 | |
clayg | peluse: acoles_away: twice now reading through this diff I've thought that it seems like we should just have storage policies define a diskfile - you guys should probably start now building a case against it - I tend to make my mind up about stuff and with out a good argument to contrarty I'll probably have to code it up | 00:29 |
peluse | clayg, yeah that was discussed way back when for sure and we know a lot more now than we did then... | 00:36 |
openstackgerrit | Tim Burke proposed openstack/python-swiftclient: Use proper checksum algorithm for large objects https://review.openstack.org/161043 | 00:38 |
clayg | I think we originally didn't iagmine all this fragment archive and .durable stuff smushing all up against the object's storage path | 00:38 |
peluse | but I dunno... something we should talk about more but unfortunately I have nasty business travel (back to the freezing f'ing cold near tdasilva) so the rest of my week is nearly shot to hell | 00:38 |
clayg | lol | 00:38 |
tdasilva | haha | 00:39 |
tdasilva | are you headed to Boston? | 00:39 |
peluse | is it snowing up there? - yes Boston | 00:39 |
peluse | leave tomroor like 10 min after the IRC meeting | 00:39 |
tdasilva | yeah, 2-3 inches overnight, another 4 inches on thu | 00:40 |
tdasilva | the thursday snow should break the record for all time high for boston, so you will be here to experience history in the making :D | 00:40 |
peluse | sheeeeeeeeit | 00:41 |
tdasilva | peluse: are you attending Vault conference next week? | 00:45 |
*** nexusz99 has joined #openstack-swift | 00:45 | |
mattoliverau | ho: morning (sorry for late response) | 00:46 |
peluse | thought about it but too much going on and I probably would have gotten a "no" anyhow with the summit coming up soon | 00:46 |
*** gyee has quit IRC | 00:51 | |
*** jrichli has joined #openstack-swift | 00:51 | |
*** gyee has joined #openstack-swift | 01:04 | |
*** annegentle has quit IRC | 01:08 | |
tdasilva | clayg: re: the issue with etag that you found, I'm changing it so that the proxy returns the correct etag in the response to the client | 01:09 |
openstackgerrit | paul luse proposed openstack/swift: Ssync fix for EC PUT path https://review.openstack.org/161057 | 01:11 |
peluse | tdasilva, BTW, feature/ec is working OK for me now that I fixed my env (eventlet) | 01:11 |
*** annegentle has joined #openstack-swift | 01:12 | |
tdasilva | yeah...that was my problem too...once you mentioned eventlet and I looked at tushars patch and noticed that it fixed exactly the issue I was running into: send_hundred_continue_response | 01:13 |
tdasilva | peluse: ^ | 01:13 |
peluse | cool | 01:13 |
clayg | tdasilva: well good! | 01:15 |
peluse | i gotta go get some grub..... | 01:16 |
*** nexusz99 has quit IRC | 01:16 | |
*** annegentle has quit IRC | 01:19 | |
*** cebruns has joined #openstack-swift | 01:38 | |
openstackgerrit | Thiago da Silva proposed openstack/swift: WIP: Use Container override header to update etag and size https://review.openstack.org/161030 | 01:41 |
clayg | redbo: I was wondering if as alternative to implementing a python vm in go so you can update and sync hashes.pkl - if we could move to a more portable format? | 01:43 |
redbo | Oh I'm off the python vm. I implemented pickles in go. | 01:46 |
clayg | redbo: I know you were thinking about doing a per-part sqlite database - but before that if the REPLICATE verb would send json, and there was some sort of upgrade path for hash invalidation in the object server from a pickle to... idk - berkley db? seems like you get get all the pickles out of your go's | 01:46 |
clayg | oh... well if that works i suppose pickle it is | 01:46 |
redbo | There's also https://github.com/hydrogen18/stalecucumber but it's kind of slow. | 01:47 |
clayg | still i wonder if a there's a better serialization format for updating key's in a dictionary than a deserialize/modify/serialize pickle | 01:47 |
clayg | oh nice | 01:48 |
clayg | ok then - sounds like pickle is gunna turn out to be a portable data format after all - wtg standards evolving from implementation! | 01:48 |
redbo | Well just adding a log to that will help a lot. Like append suffixes to a file that says "next time you recalculate hashes, invalidate this one". | 01:49 |
redbo | That will lighten the load for PUTs and DELETEs | 01:49 |
redbo | and push it on replication where it belongs | 01:50 |
clayg | hashes.pkl.pending ? | 01:50 |
clayg | :P | 01:50 |
redbo | :) | 01:52 |
clayg | so are you rewriting the replicators in go too? or did you get nerd sniped trying to get feature parity in the proxy? | 01:53 |
redbo | We're playing around with a replicator. I don't think we'll ever get a full proxy, but if we could get a simple proxy that serves a couple of our big use cases (like serving CDN traffic) it'd be interesting. | 01:57 |
clayg | i think i like the .pending idea - if the object server could just start appending invalidates on writes - you could convert the hashes data format however you might like during replication | 01:57 |
redbo | oh, true. But I'd like it to append "add file X/remove file Y" :) | 01:58 |
clayg | sounds awesome, keep me posted, let me know if there's anything I can do to help (besides... you know... breaking handoff partition cleanup :\) | 01:58 |
clayg | redbo: one thing at a time? | 01:58 |
clayg | :P | 01:58 |
*** jamielennox is now known as jamielennox|away | 01:59 | |
*** reed has quit IRC | 02:00 | |
clayg | oh.. no, that was *one* thing - instead of append "suff x, suff x, suff x" over and over you'd go ahead and put the name an op in there - but then still do the "apply .pending" as read all the objects, squash them down to the list of suffix to invalidate, and update the hashes.pkl | 02:00 |
clayg | redbo: you have to go more slowly with me | 02:00 |
redbo | haha yeah... just to future proof it. Okay, $lady is here for dinner. | 02:02 |
openstackgerrit | Samuel Merritt proposed openstack/swift: Fix EC chunking https://review.openstack.org/161070 | 02:02 |
*** dmsimard_away is now known as dmsimard | 02:08 | |
redbo | I mean.. don't tell her she's a variable and not a defined constant. | 02:11 |
*** timburke has joined #openstack-swift | 02:13 | |
*** nexusz99 has joined #openstack-swift | 02:17 | |
*** annegentle has joined #openstack-swift | 02:20 | |
*** nexusz99 has quit IRC | 02:21 | |
clayg | oh my | 02:24 |
clayg | timburke: welcome to the shit! https://review.openstack.org/#/c/161043/ | 02:25 |
*** annegentle has quit IRC | 02:25 | |
clayg | ho: mattoliverau: who were the guys talking about slo md5's at the hack-a-thon? | 02:25 |
ho | clayg: miyahara-san, I think. I will check first name. Please wait... | 02:26 |
ho | clayg: his first name is Kazuhiro | 02:27 |
mattoliverau | clayg: yeah what ho said.. but I dont know his irc nick | 02:27 |
mattoliverau | dmorita: what's Kazuhiro irc nick ^^ | 02:27 |
*** jamielennox|away is now known as jamielennox | 02:37 | |
dmorita | mattoliverau: sorry i do not know his nick but i can forward a message to him :) | 02:37 |
mattoliverau | dmorita: no probs :) | 02:38 |
dmorita | Ahh, patch 161043 is relating to Kazuhiro's suggestion! I will tell him this patch. | 02:42 |
clayg | dmorita: thanks man! | 02:49 |
*** jamielennox is now known as jamielennox|away | 02:52 | |
*** jamielennox|away is now known as jamielennox | 02:55 | |
*** jamielennox is now known as jamielennox|away | 02:56 | |
*** jamielennox|away is now known as jamielennox | 02:57 | |
*** annegentle has joined #openstack-swift | 03:05 | |
*** annegentle has quit IRC | 03:07 | |
*** nexusz99 has joined #openstack-swift | 03:09 | |
*** gyee has quit IRC | 03:29 | |
*** tsg has joined #openstack-swift | 03:30 | |
clayg | peluse: ok, i read over 159637 - i'm trying to think of how to exercise some of it functionaly... | 03:43 |
clayg | maybe i write some scripts to play around with the suffix pickles and replication requests... maybe if I nuter _finalize_put so it doesn't call hash_cleanup_listdir I can get some interesting hashdirs | 03:44 |
*** jrichli has quit IRC | 03:47 | |
*** nexusz99 has quit IRC | 04:00 | |
*** haomaiwang has joined #openstack-swift | 04:04 | |
*** ho__ has joined #openstack-swift | 04:06 | |
*** ho has quit IRC | 04:06 | |
*** dmsimard is now known as dmsimard_away | 04:29 | |
*** ho has joined #openstack-swift | 04:34 | |
*** ho__ has quit IRC | 04:35 | |
*** openstackgerrit has quit IRC | 04:46 | |
*** kkkkkkk has joined #openstack-swift | 04:49 | |
*** openstackgerrit has joined #openstack-swift | 04:53 | |
*** bkopilov has joined #openstack-swift | 05:05 | |
*** kkkkkkk has quit IRC | 05:08 | |
*** ppai has joined #openstack-swift | 05:17 | |
*** zaitcev has quit IRC | 05:24 | |
*** jamielennox is now known as jamielennox|away | 05:32 | |
*** openstack has joined #openstack-swift | 05:36 | |
openstackgerrit | Hisashi Osanai proposed openstack/swift: Enable Object Replicator's failure count in recon https://review.openstack.org/138342 | 05:46 |
*** silor has joined #openstack-swift | 06:05 | |
*** bsdkurt has quit IRC | 06:14 | |
*** Sanchit has quit IRC | 06:17 | |
*** bsdkurt has joined #openstack-swift | 06:25 | |
*** SkyRocknRoll has joined #openstack-swift | 06:39 | |
*** btorch has quit IRC | 06:45 | |
*** eikke has quit IRC | 07:19 | |
openstackgerrit | Prashanth Pai proposed openstack/swift: fsync() on directories https://review.openstack.org/126923 | 07:25 |
*** david-lyle_afk has joined #openstack-swift | 07:33 | |
*** openstackgerrit has quit IRC | 07:49 | |
*** openstackgerrit has joined #openstack-swift | 07:50 | |
*** zacksh has quit IRC | 07:52 | |
*** mmcardle has joined #openstack-swift | 07:59 | |
*** rledisez has joined #openstack-swift | 08:17 | |
*** zacksh has joined #openstack-swift | 08:17 | |
*** chlong has quit IRC | 08:20 | |
*** geaaru has joined #openstack-swift | 08:22 | |
*** mmcardle has quit IRC | 08:25 | |
*** mmcardle has joined #openstack-swift | 08:29 | |
-openstackstatus- NOTICE: Zuul check queue stuck due to reboot maintenance window at one of our cloud providers - no need to recheck changes at the moment, they won't move forward. | 08:32 | |
*** ChanServ changes topic to "Zuul check queue stuck due to reboot maintenance window at one of our cloud providers - no need to recheck changes at the moment, they won't move forward." | 08:32 | |
*** jordanP has joined #openstack-swift | 08:48 | |
*** jistr has joined #openstack-swift | 08:50 | |
*** ho has quit IRC | 08:58 | |
*** km has quit IRC | 09:04 | |
*** openstack has joined #openstack-swift | 15:28 | |
*** btorch has joined #openstack-swift | 15:38 | |
*** mattoliverau has quit IRC | 15:43 | |
*** eikke has joined #openstack-swift | 16:03 | |
*** Bsony has quit IRC | 16:03 | |
*** mattoliverau has joined #openstack-swift | 16:05 | |
notmyname | good morning | 16:07 |
tdasilva | notmyname: morning! :) | 16:12 |
*** tsg has quit IRC | 16:12 | |
mahatic | good morning! | 16:14 |
openstackgerrit | Stuart McLaren proposed openstack/python-swiftclient: Allow reading from object body on download https://review.openstack.org/155291 | 16:14 |
openstackgerrit | Stuart McLaren proposed openstack/python-swiftclient: Retry download of object body https://review.openstack.org/159208 | 16:14 |
openstackgerrit | Thiago da Silva proposed openstack/swift: WIP: Use Container override header to update etag and size https://review.openstack.org/161030 | 16:15 |
notmyname | mahatic: feeling better today? | 16:15 |
mahatic | notmyname, yup, quite a bit. thanks for checking :) | 16:16 |
notmyname | timburke: you submitted a patch! awesome! | 16:16 |
* notmyname didn't even know you were working on one :-) | 16:17 | |
notmyname | looks like the community qa cluster is back online. yay | 16:20 |
*** annegentle has quit IRC | 16:21 | |
*** annegentle has joined #openstack-swift | 16:22 | |
*** rdaly2_ has joined #openstack-swift | 16:25 | |
*** rdaly2 has quit IRC | 16:29 | |
*** mahatic has quit IRC | 16:42 | |
*** mrda has quit IRC | 16:43 | |
*** mahatic has joined #openstack-swift | 16:44 | |
*** btorch has quit IRC | 16:52 | |
*** btorch has joined #openstack-swift | 16:52 | |
notmyname | mahatic: yesterday I think you said you had some questions? | 17:02 |
notmyname | eikke: so I see that you're working on some py26 compat issues in your diskfile adapter. how important is py26 to you? | 17:04 |
mahatic | notmyname, yeah. I'm still looking actually (at the patch). I'm doing some more clarifications. Can I ping you back in a bit? | 17:06 |
notmyname | eikke: I ask because there's been some talk about dropping py26 compat post-kilo. most recently because of https://bugs.launchpad.net/swift/+bug/1425776 | 17:06 |
openstack | Launchpad bug 1425776 in OpenStack Object Storage (swift) "ACL tests fails on fresh Ubuntu 14.04" [Undecided,In progress] - Assigned to Christian Schwede (cschwede) | 17:06 |
notmyname | mahatic: sure | 17:07 |
mahatic | thanks. | 17:07 |
*** tsg has joined #openstack-swift | 17:12 | |
*** jistr has quit IRC | 17:13 | |
notmyname | tsg: good morning! (I have questions for you) | 17:15 |
-openstackstatus- NOTICE: Issue solved, gate slowly digesting accumulated changes | 17:17 | |
*** nellysmitt has quit IRC | 17:18 | |
*** mmcardle has quit IRC | 17:23 | |
tsg | notmyname: morning! how is it going | 17:26 |
notmyname | tsg: good | 17:26 |
notmyname | tsg: as we're getting closer to EC being done, getting pyeclib available on pypi is getting more important. so 2 questions: | 17:27 |
notmyname | tsg: when will the current (1.0RC?) be available in pypi? | 17:28 |
notmyname | tsg: I'm told that the current version on pypi (0.9.8) still does external calls on install. is that true? if so, when will that be fixed? | 17:28 |
tsg | notmyname: to get pypi version up-to-date, there is a dependency on liberasurecode - that needs to be added to jenkins slaves or else pyeclib 1.0 will not work. liberasurecode is available as a Debian package now - I have been trying to understand from zigo what's needed from a package availability standpoint | 17:30 |
notmyname | got it. thanks | 17:31 |
tsg | if the package has to be pulled into ubuntu repos to be a part of jenkins slaves .. | 17:31 |
notmyname | tsg: and thanks for working on it | 17:31 |
mahatic | notmyname, I have a basic question: Shouldn't the ring take care of all the servers no matter which storage policy? | 17:32 |
notmyname | tsg: I'd love to see some sort of resolution this week so that there are fewer barriers for testing the EC work | 17:32 |
tsg | notmyname: no problem - zigo not being reachable for the past few days has made it go a bit slow | 17:32 |
tsg | notmyname: trying my best | 17:32 |
tsg | will talk to clarkb on some alternatives | 17:32 |
notmyname | mahatic: there's a different object ring for each storage policy. I'd love to see swift-recon be able to identify the storage policy and then it will pick up the appropriate ring | 17:33 |
notmyname | tsg: thanks | 17:33 |
mahatic | oh okay | 17:33 |
notmyname | mahatic: and it looks like ho and dmorita are working on some recon updates (they're both in japan and not online right now) | 17:34 |
mahatic | notmyname, https://review.openstack.org/#/c/138697/ This is one is in progress only for auditor-check I believe? | 17:34 |
notmyname | mahatic: ya, there are actually 2 parts to policy-aware recon | 17:35 |
mahatic | notmyname, oh | 17:35 |
notmyname | mahatic: one, like dmorita is working on there, is to have the recon middleware report back on info found for any storage policy on a given host | 17:35 |
notmyname | mahatic: the other is on the CLI side knowing which host to actually query. that's where the different rings come into play | 17:36 |
notmyname | does that make sense? | 17:37 |
mahatic | notmyname, okay. But through the first part you mentioned, it would also report the host right? | 17:38 |
*** rledisez has quit IRC | 17:38 | |
notmyname | I don't follow what you mean | 17:39 |
mahatic | notmyname, yeah, my question was wrong. I sort of understood what you said. But I do see a huge gap in "finding out which host to query" | 17:41 |
notmyname | I think with dmorita's patch right now it will only find the per-policy auditor info for hosts that happen to be in the default storage policy 0 and also in some other policy | 17:42 |
notmyname | but if there are hosts that are not in the legacy storage policy at all, then swift-recon has no way to query them | 17:43 |
notmyname | I think | 17:43 |
mahatic | I think i'm really missing something. Is it not that, you can only have one storage policy as a default? | 17:43 |
mahatic | oh | 17:44 |
mahatic | no, object can have one, container can have another policy, etc correct? | 17:44 |
notmyname | no | 17:45 |
notmyname | policies are only for objects | 17:45 |
notmyname | but the "legacy" policy is storage policy 0 (that's what I meant by default. sorry that was the wrong word) | 17:45 |
*** gyee has joined #openstack-swift | 17:47 | |
*** Nadeem has joined #openstack-swift | 17:47 | |
mahatic | okay. the thing you said about dmorita's patch makes sense. | 17:48 |
notmyname | mahatic: so as I see it right now, there's not a way to do a `swift-recon` with anything other than storage policy 0. ie the one with the object.ring.rz. you can't load and query hosts that are referenced in object-N.ring.gz | 17:48 |
acoles | peluse: clayg: around? | 17:49 |
mahatic | right | 17:49 |
mahatic | notmyname, I understand on a higher level, but not so much on the technical one. I'll need to look. But I got what you mean | 17:50 |
notmyname | mahatic: then you're at the same place as me :-) | 17:51 |
mahatic | lol | 17:51 |
mahatic | notmyname, well, I'll try to get ahead of you on that :) | 17:52 |
mahatic | notmyname, what I also meant was, I haven't really looked at how rings for different policies are done and/or accessed. I'll start from there. | 17:53 |
*** erlon has joined #openstack-swift | 18:00 | |
mahatic | notmyname, This is sure to keep me busy tomorrow! I'm calling it a night for now. Will check the meeting log later. | 18:01 |
notmyname | :-) | 18:01 |
notmyname | so at this week's team meeting (in 1 hour), we'll discuss EC progress. and we should make a call on if we are committed to an EC beta in Kilo or not | 18:04 |
mahatic | yeah, gotta sleep before my running nose gets the better of me :D | 18:04 |
*** mahatic has quit IRC | 18:06 | |
*** ChanServ changes topic to "Review Dashboard: http://goo.gl/uRzLBX | Overview Dashboard: http://goo.gl/9EI0Sz | Priority Reviews: https://wiki.openstack.org/wiki/Swift/PriorityReviews | Ideas: https://wiki.openstack.org/wiki/Swift/ideas" | 18:07 | |
tsg | notmyname: just talked to clarkb on -infra. It appears that the package has to be a part of the standard Ubuntu repos. | 18:09 |
tsg | notmyname: typically Debian packages are syned to Ubuntu on every (dot) release - doesn't look like liberasurecode was synced on 14.04.2 (that went out on 02/19) though. I just filed an Ubuntu bug to sync the package. | 18:09 |
notmyname | hmm | 18:09 |
notmyname | tsg: so what if it's not? does that mean we can't use pyeclib and therefore can't test any ec changes in the gate? | 18:10 |
notmyname | tsg: also, that makes me wonder what ec testing is on the feature/ec branch today | 18:10 |
tsg | notmyname: 0.9.8 is working well right now | 18:10 |
notmyname | ok | 18:10 |
tsg | we are all testing using 0.9.8 | 18:10 |
tsg | (and Kevin/I/Kota with 1.0 also, ofcourse) | 18:11 |
notmyname | tsg: ah, ok. so 0.9.8 is in pypi. and the right c library for that is in repos? it's just updated versions? | 18:11 |
notmyname | or is there no c library component in 0.9.8? | 18:11 |
tsg | notmyname: the C code is all part of pyeclib in 0.9.8, not split out into liberasurecode | 18:12 |
notmyname | oh ok | 18:12 |
tsg | notmyname: that was done on Debian/RH guys' request | 18:12 |
tsg | plus it was also good to create a generic C API library as well | 18:13 |
notmyname | ok | 18:13 |
notmyname | tsg: and the question around the external calls? | 18:13 |
notmyname | during install | 18:13 |
tsg | that's the other thing we changed in liberasurecode | 18:13 |
notmyname | the split of the c library has the change that removes the external calls at install? | 18:14 |
tsg | we designed it with dl so that the compilation will not require pulling all the dependencies of the Internet | 18:14 |
tsg | yes | 18:14 |
notmyname | ok | 18:14 |
notmyname | what's "dl"? | 18:14 |
tsg | dynamic loader | 18:15 |
notmyname | ah ok | 18:15 |
notmyname | tsg: how worried should I be about this impacting an EC release schedule? :-) | 18:15 |
notmyname | (cause it sounds troublesome) | 18:15 |
tsg | notmyname: because the Debian package for liberasurecode and jerasure is already, it shouldn't be | 18:17 |
tsg | I will find out asap | 18:17 |
*** logan2 has quit IRC | 18:18 | |
notmyname | tsg: is my understanding correct? EC in swift will require pyeclib 1.0 and 0.9.8 will be unsuitable. and therefore the risk is that it will not be synced from debian to ubuntu in time? | 18:18 |
notmyname | (/me thinks the remote loading in 0.9.8 is "unsuitable") | 18:19 |
tsg | notmyname: hm clarkb didn't seem completely against it | 18:19 |
*** jordanP has quit IRC | 18:20 | |
tsg | notmyname: but yes, we should move to 1.0 asap given the API changes/fixes it brings | 18:20 |
notmyname | tsg: ok | 18:20 |
notmyname | tsg: we have 4 weeks to get this packaging/dependency stuff resolved, if we are going to make kilo. do you think that's reasonable? | 18:22 |
notmyname | tsg: oh, actually less than that. that is for our swift code. the feature freeze is before that. let me find that date | 18:22 |
tsg | notmyname: I think so, if I can get hold of zigo :) | 18:23 |
tsg | notmyname: I will try to talk to ubuntu folks today | 18:23 |
notmyname | march 19 for the openstack dependency freeze | 18:23 |
notmyname | so 15 calendar days | 18:23 |
notmyname | 2 weeks | 18:23 |
notmyname | tsg: pyeclib 0.9.8 is in global requirements right now and installed on the openstack CI boxes? | 18:26 |
tsg | notmyname: yes - it is installed on CI slaves | 18:26 |
notmyname | ok, thanks | 18:26 |
tsg | (what our EC tests have been using) | 18:27 |
tsg | brb | 18:27 |
*** tsg has left #openstack-swift | 18:27 | |
*** tsg has joined #openstack-swift | 18:27 | |
*** geaaru has quit IRC | 18:28 | |
*** acoles is now known as acoles_away | 18:29 | |
*** zhill has joined #openstack-swift | 18:34 | |
openstackgerrit | Thiago da Silva proposed openstack/swift: Use Container override header to update etag and size https://review.openstack.org/161030 | 18:48 |
*** gyee has quit IRC | 18:49 | |
*** tab____ has joined #openstack-swift | 18:50 | |
*** cutforth has joined #openstack-swift | 18:52 | |
*** rdaly2_ has quit IRC | 18:55 | |
*** kota_ has joined #openstack-swift | 18:57 | |
*** zhill_ has joined #openstack-swift | 18:58 | |
mattoliverau | morning | 18:58 |
notmyname | mattoliverau: hello | 18:58 |
kota_ | morning | 18:59 |
notmyname | mattoliverau: thanks for waking up early :-) | 18:59 |
notmyname | kota_: wow. you're up really early too! | 18:59 |
notmyname | more than mattoliverau | 18:59 |
kota_ | notmyname: :) | 18:59 |
mattoliverau | yeah, go kota_ | 18:59 |
zaitcev | TZ=Japan date Thu Mar 5 03:59:57 JST 2015 | 19:00 |
notmyname | ok, meeting time in #openstack-meeting | 19:00 |
zaitcev | taihensoudesu | 19:00 |
kota_ | zaitcev: no worries :) | 19:01 |
notmyname | clayg: torgomatic: acoles_away: cschwede: dfg: redbo: meeting ping | 19:01 |
notmyname | peluse: ^ | 19:02 |
*** zhill_ has quit IRC | 19:03 | |
notmyname | tsg: ^ | 19:03 |
*** acoles_away is now known as acoles | 19:05 | |
*** lpabon has joined #openstack-swift | 19:15 | |
*** monn has joined #openstack-swift | 19:15 | |
*** gvernik has joined #openstack-swift | 19:17 | |
*** ChanServ changes topic to "Hackathon: https://www.eventbrite.com/e/openstack-swift-hack-a-thon-sponsored-by-swiftstack-tickets-15226147818 | Priority Reviews: https://wiki.openstack.org/wiki/Swift/PriorityReviews | Review Dashboard: http://goo.gl/r2mxbe" | 19:17 | |
*** gyee has joined #openstack-swift | 19:19 | |
clayg | is that now? | 19:21 |
tdasilva | y | 19:22 |
acoles | clayg: yup :D | 19:22 |
*** Bsony has joined #openstack-swift | 19:24 | |
*** aix has quit IRC | 19:24 | |
*** gvernik has quit IRC | 19:25 | |
acoles | clayg: still here? | 19:32 |
notmyname | meeting notes: http://eavesdrop.openstack.org/meetings/swift/2015/swift.2015-03-04-19.00.html | 19:32 |
clayg | acoles: sure | 19:32 |
notmyname | oh, note that the review dashboard links in the topic have a "starred" section for priority reviews. I meant to mention that in the meeting | 19:33 |
acoles | clayg: hi. was saying in meeting - i can spend more time on https://review.openstack.org/#/c/159637 but saw your comment in scrollback about refactoring into per-policy diskfiles | 19:33 |
acoles | clayg: so.... am i wasting my time? peluse thinks not | 19:34 |
acoles | clayg: fwiw i see where you are coming from, there's plenty of 'if this policy or that' | 19:34 |
acoles | some of which i managed to puch out into plicy classes | 19:35 |
acoles | notmyname: cool | 19:35 |
clayg | acoles: i'm debating this with myself | 19:35 |
clayg | acoles: there's a lot of code in the ec diskfile that will basically be exactly the same but do this one extra little thing different | 19:35 |
acoles | clayg: yup | 19:36 |
clayg | acoles: in some respects I think it'd be better to just duplicate it - in others I think we'll find ourselves wanting to make the same changes in two places :'( | 19:36 |
clayg | acoles: either way - I think we should keep hacking on that patch | 19:36 |
clayg | acoles: I'm going to try and contribute by cleaning up the abstraction of the new suffix class thing - based on previous ideas about improving object replciation it seems like we could move that abstraction around to make it easier to 1) do a different thing for ec 2) make it easier to do a different thing for replication | 19:37 |
acoles | clayg: ok, i have not looked at the suffix class, so no overlap there | 19:38 |
acoles | clayg: isit worth splitting the patch up into smaller pieces or is it all inter-dependent? | 19:38 |
clayg | acoles: my experience reiewing proxy changes to accomodate ec may be coloring my view - in the proxy the ec path is much different than in the object server path | 19:38 |
clayg | acoles: I think peluse is already trying to split it up | 19:38 |
acoles | clayg: yeah, don't get me wrong, i sympathise, just wondering what best way to use my effort is | 19:39 |
clayg | acoles: I also think getting something merged onto feature/ec is a higher priority than trying to split up the change for master | 19:39 |
acoles | clayg: agree | 19:39 |
clayg | er... rather we'll have to rethink how to present the logical changes when we push to master - so for now I'm more interested in just running as fast as we can to "works" on feature/ec | 19:39 |
clayg | if it changes before master it won't make a difference - the master change log will only show the final state | 19:40 |
acoles | clayg: ok so a plan: tomorrow i will fix get_ondisk_files and hash_cleanup some more, and look for othe rplaces where ec policy behavior can be encapsulated into policy class, and where we can resolve policy and stop passing around the policy_index | 19:41 |
*** kota_ has quit IRC | 19:41 | |
acoles | clayg: i'll leave suffix hash stuff alone. | 19:41 |
acoles | clayg: that sound ok? | 19:41 |
clayg | acoles: sounds great, thanks for the help! | 19:43 |
acoles | clayg: cool, i love a plan! | 19:43 |
acoles | jrichli: i reviewed your middleware patch, but as an afterthought, maybe focus on the en/decrypter pieces for now and not worry about trivialkeymaster. Thats easily separable. | 19:44 |
jrichli | acoles: understood | 19:46 |
jrichli | acoles: and thanks for reviewing! I will go over the comments now. | 19:46 |
*** ChanServ sets mode: +v mattoliverau | 19:47 | |
acoles | jrichli: np. have a good afternoon, i'm done for today. | 19:47 |
jrichli | have a good evening! | 19:47 |
*** acoles is now known as acoles_away | 19:47 | |
mattoliverau | night acoles! | 19:48 |
mattoliverau | or now acoles_away ^^ | 19:48 |
*** Bsony has quit IRC | 19:48 | |
*** EmilienM is now known as EmilienM|afk | 19:56 | |
*** reed has joined #openstack-swift | 19:56 | |
*** devlaps has joined #openstack-swift | 19:58 | |
*** tsg has quit IRC | 20:10 | |
*** tsg has joined #openstack-swift | 20:10 | |
*** rdaly2 has joined #openstack-swift | 20:10 | |
*** tsg has quit IRC | 20:16 | |
clayg | peluse: I don't think that tombstones really have #frag_index parts in their filename? | 20:23 |
openstackgerrit | OpenStack Proposal Bot proposed openstack/swift: Updated from global requirements https://review.openstack.org/88736 | 20:23 |
*** logan2 has joined #openstack-swift | 20:29 | |
*** chlong has joined #openstack-swift | 20:29 | |
notmyname | well, now it's real. I just sent out a status update to ${not_swift_devs} about having an EC beta in Kilo | 20:38 |
notmyname | grrr...looks like the openstack bot resurrected a stale topic message | 20:39 |
openstackgerrit | Samuel Merritt proposed openstack/swift: Fix EC chunking https://review.openstack.org/161070 | 20:39 |
*** ChanServ changes topic to "Review Dashboard: http://goo.gl/uRzLBX | Overview Dashboard: http://goo.gl/9EI0Sz | Priority Reviews: https://wiki.openstack.org/wiki/Swift/PriorityReviews | Ideas: https://wiki.openstack.org/wiki/Swift/ideas" | 20:40 | |
*** dmsimard is now known as dmsimard_away | 20:53 | |
*** Bsony has joined #openstack-swift | 20:55 | |
*** silor has quit IRC | 20:56 | |
*** Bsony has quit IRC | 20:59 | |
clayg | notmyname: is there any progress on py26 as a gate check? do we run our tests on the feature/ec branch? assertGreaterEqual :'( | 21:14 |
notmyname | clayg: there is no py26 check in the official gate. (and there will not be). I've mentioned it to charz to add it to the community cluster testing | 21:15 |
notmyname | I just saw this am that the community cluster is back to doing tests and reporting. so I think that can be added to it | 21:15 |
clayg | notmyname: yeah that's what I ment | 21:16 |
notmyname | so now that the cluster is back online, I'll follow up with him asap | 21:17 |
*** tsg has joined #openstack-swift | 21:18 | |
*** tsg_ has joined #openstack-swift | 21:20 | |
*** Anticimex has quit IRC | 21:21 | |
*** sandywalsh has quit IRC | 21:23 | |
*** tsg has quit IRC | 21:23 | |
*** sandywalsh has joined #openstack-swift | 21:24 | |
*** Anticimex has joined #openstack-swift | 21:27 | |
*** rdaly2 has quit IRC | 21:29 | |
*** monn has quit IRC | 21:31 | |
*** EmilienM|afk is now known as EmilienM | 21:43 | |
*** tsg_ has quit IRC | 22:02 | |
*** tsg_ has joined #openstack-swift | 22:03 | |
*** tsg_ has quit IRC | 22:08 | |
*** annegentle has quit IRC | 22:10 | |
*** annegentle has joined #openstack-swift | 22:11 | |
* notmyname goes offline for a while | 22:12 | |
*** chlong has quit IRC | 22:12 | |
*** jamielennox|away is now known as jamielennox | 22:14 | |
*** erlon has quit IRC | 22:41 | |
*** Anticimex has quit IRC | 22:49 | |
*** tsg has joined #openstack-swift | 22:50 | |
*** Anticimex has joined #openstack-swift | 22:51 | |
clayg | notmyname: not sure if you're around - but something is going on with the gate not testing python-swiftclient patches maybe? | 23:00 |
*** annegentle has quit IRC | 23:04 | |
clayg | nm, turns out if you scroll all the way down to the bottom of http://status.openstack.org/zuul/ the graphs tell the story - bunch of vms got turned off last night at rax and there's some issues getting new ones up | 23:07 |
*** jrichli has quit IRC | 23:10 | |
*** tsg_ has joined #openstack-swift | 23:17 | |
*** tsg has quit IRC | 23:19 | |
*** annegentle has joined #openstack-swift | 23:25 | |
*** annegentle has quit IRC | 23:32 | |
*** annegentle has joined #openstack-swift | 23:33 | |
clayg | notmyname: you back? my +3 on https://review.openstack.org/#/c/161070/ didn't work, it's still not merged | 23:38 |
clayg | acoles_away: peluse: cschwede: mattoliverau: zaitcev: redbo: dfg: glange: ^ | 23:41 |
clayg | swifterdarrell: ^ | 23:41 |
*** Nadeem has quit IRC | 23:41 | |
*** tab____ has quit IRC | 23:42 | |
*** chlong has joined #openstack-swift | 23:42 | |
*** km has joined #openstack-swift | 23:42 | |
*** annegentle has quit IRC | 23:43 | |
*** annegentle has joined #openstack-swift | 23:43 | |
*** chlong has quit IRC | 23:46 | |
*** chlong has joined #openstack-swift | 23:46 | |
*** annegentle has quit IRC | 23:48 | |
mattoliverau | clayg: will take a loom | 23:49 |
mattoliverau | Look even | 23:50 |
*** Tahmina has joined #openstack-swift | 23:50 | |
* clayg hugs mattoliverau | 23:50 | |
mattoliverau | Aww | 23:52 |
*** EmilienM is now known as EmilienM|afk | 23:56 | |
*** annegentle has joined #openstack-swift | 23:56 | |
*** annegentle has quit IRC | 23:57 | |
*** annegentle has joined #openstack-swift | 23:57 | |
*** annegentle has quit IRC | 23:58 | |
*** dmsimard_away is now known as dmsimard | 23:58 | |
openstackgerrit | Tim Burke proposed openstack/python-swiftclient: Compare each chunk of large objects when uploading https://review.openstack.org/161043 | 23:58 |
*** Tahmina has quit IRC | 23:58 | |
*** dmsimard is now known as dmsimard_away | 23:58 |
Generated by irclog2html.py 2.14.0 by Marius Gedminas - find it at mg.pov.lt!