| openstackgerrit | Matthew Oliver proposed openstack/swift: Correct ringbuilder's set_weight usage string https://review.openstack.org/426620 | 00:08 |
|---|---|---|
| notmyname | timburke: I'm not sure what this is in reference to, but it sounds related to the auth v1 keystone plugin you made http://lists.openstack.org/pipermail/openstack/2017-January/018472.html | 00:12 |
| *** NM has joined #openstack-swift | 01:24 | |
| *** foexle has quit IRC | 01:35 | |
| *** sams-gleb has joined #openstack-swift | 01:35 | |
| *** sams-gleb has quit IRC | 01:40 | |
| kota_ | hello world | 02:00 |
| kota_ | back to work shortly from my sick. | 02:00 |
| mattoliverau | kota_: morning, welcome back! | 02:08 |
| kota_ | mattoliverau: thanks! | 02:15 |
| kota_ | mattoliverau: however, i'm planning to work short today, it doesn't seem still completely fine yet. | 02:18 |
| mattoliverau | kota_: well it's good to see that your starting to feel much better! but don't work too hard, it's better to rest and get fully better rather then be not quite 100% for a long time. | 02:19 |
| *** tanee_away is now known as tanee | 02:31 | |
| *** NM has quit IRC | 02:31 | |
| *** rcernin has joined #openstack-swift | 03:11 | |
| *** rcernin has quit IRC | 03:13 | |
| *** rcernin has joined #openstack-swift | 03:14 | |
| *** rcernin has quit IRC | 03:25 | |
| *** sams-gleb has joined #openstack-swift | 03:38 | |
| *** sams-gleb has quit IRC | 03:42 | |
| *** SkyRocknRoll has joined #openstack-swift | 03:49 | |
| *** links has joined #openstack-swift | 03:56 | |
| *** links has quit IRC | 03:57 | |
| mahatic | good morning | 04:12 |
| mahatic | notmyname: +1 on your comment on the US news | 04:14 |
| mahatic | kota_: good to know you're feeling better | 04:15 |
| kota_ | mahatic: good morning, and thanks | 04:15 |
| *** maestropandy has joined #openstack-swift | 04:31 | |
| *** vinsh has joined #openstack-swift | 04:31 | |
| mattoliverau | mahatic: morning | 04:32 |
| mahatic | mattoliverau: o/ | 04:32 |
| *** vinsh has quit IRC | 04:51 | |
| *** klrmn has quit IRC | 05:04 | |
| *** psachin has joined #openstack-swift | 05:28 | |
| *** maestropandy has left #openstack-swift | 05:33 | |
| *** sams-gleb has joined #openstack-swift | 05:40 | |
| *** sams-gleb has quit IRC | 05:45 | |
| *** takashi has joined #openstack-swift | 05:57 | |
| openstackgerrit | Kota Tsuyuzaki proposed openstack/swift: Optimize ec duplication and its md5 hashing https://review.openstack.org/421673 | 06:01 |
| openstackgerrit | Kota Tsuyuzaki proposed openstack/swift: EC Fragment Duplication - Foundational Global EC Cluster Support https://review.openstack.org/219165 | 06:01 |
| kota_ | Update only my priority patches, I know I have a stack of incoming reviews but I'd like to step by small so pend them for the rest of this week. Getting back home to take a rest/care of my family. | 06:04 |
| * kota_ is in baby steps | 06:04 | |
| *** ppai has joined #openstack-swift | 06:17 | |
| *** takashi has quit IRC | 07:02 | |
| *** rcernin has joined #openstack-swift | 07:11 | |
| *** rcernin has quit IRC | 07:12 | |
| *** rcernin has joined #openstack-swift | 07:12 | |
| mattoliverau | kota_: yeah, don't rush. you'll be back to full speed in no time :) | 07:13 |
| *** silor has joined #openstack-swift | 07:21 | |
| *** silor1 has joined #openstack-swift | 07:34 | |
| *** silor has quit IRC | 07:36 | |
| *** silor1 is now known as silor | 07:36 | |
| *** sams-gleb has joined #openstack-swift | 07:42 | |
| *** sams-gleb has quit IRC | 07:47 | |
| *** takashi has joined #openstack-swift | 07:47 | |
| *** tesseract has joined #openstack-swift | 07:50 | |
| *** ChubYann has quit IRC | 08:00 | |
| *** sams-gleb has joined #openstack-swift | 08:08 | |
| *** rledisez has joined #openstack-swift | 08:12 | |
| *** pcaruana has joined #openstack-swift | 08:15 | |
| *** oshritf has joined #openstack-swift | 08:27 | |
| *** geaaru has joined #openstack-swift | 08:34 | |
| *** rcernin has quit IRC | 09:31 | |
| *** rcernin has joined #openstack-swift | 09:33 | |
| *** oshritf has quit IRC | 09:36 | |
| *** oshritf has joined #openstack-swift | 09:40 | |
| *** jistr has quit IRC | 09:49 | |
| *** jistr has joined #openstack-swift | 09:51 | |
| *** maestropandy has joined #openstack-swift | 09:56 | |
| *** maestropandy has quit IRC | 10:00 | |
| *** oshritf has quit IRC | 10:02 | |
| *** takashi has quit IRC | 10:05 | |
| *** maestropandy has joined #openstack-swift | 10:05 | |
| *** maestropandy has left #openstack-swift | 10:05 | |
| *** oshritf has joined #openstack-swift | 10:08 | |
| *** sams-gleb has quit IRC | 10:21 | |
| *** sams-gleb has joined #openstack-swift | 10:21 | |
| *** sams-gleb has quit IRC | 10:25 | |
| *** mvk has quit IRC | 10:28 | |
| *** sams-gleb has joined #openstack-swift | 10:37 | |
| *** mvk has joined #openstack-swift | 10:57 | |
| *** kei_yama has quit IRC | 11:00 | |
| openstackgerrit | Matthew Oliver proposed openstack/swift: Correct ringbuilder's set_weight usage string https://review.openstack.org/426620 | 11:19 |
| *** jordanP has joined #openstack-swift | 11:23 | |
| *** dims has joined #openstack-swift | 11:26 | |
| *** oshritf_ has joined #openstack-swift | 11:31 | |
| *** psachin has quit IRC | 11:32 | |
| *** oshritf has quit IRC | 11:33 | |
| *** vint_bra has joined #openstack-swift | 11:54 | |
| *** bkopilov has quit IRC | 12:02 | |
| *** foexle has joined #openstack-swift | 12:19 | |
| *** catintheroof has joined #openstack-swift | 12:25 | |
| *** NM has joined #openstack-swift | 12:33 | |
| *** bkopilov has joined #openstack-swift | 12:41 | |
| *** wasmum has quit IRC | 13:03 | |
| *** oshritf__ has joined #openstack-swift | 13:15 | |
| *** oshritf_ has quit IRC | 13:18 | |
| *** caiobrentano has joined #openstack-swift | 14:16 | |
| *** caiobrentano has quit IRC | 14:23 | |
| *** vills_ has joined #openstack-swift | 14:26 | |
| openstackgerrit | Merged openstack/swift: Correct ringbuilder's set_weight usage string https://review.openstack.org/426620 | 14:27 |
| openstackgerrit | Béla Vancsics proposed openstack/swift: Reduced the complexity of the _response_iter method https://review.openstack.org/426782 | 14:41 |
| *** rcernin has quit IRC | 14:48 | |
| *** _JZ_ has joined #openstack-swift | 14:57 | |
| *** rcernin has joined #openstack-swift | 15:00 | |
| notmyname | tdasilva: cschwede: this is interesting http://martin.kleppmann.com/2017/01/26/data-loss-in-large-clusters.html | 15:01 |
| *** mvk has quit IRC | 15:02 | |
| cschwede | notmyname: Hi! Yes, but nothing new. Summarized: the bigger the cluster, the more likely it is (as an operator) that there is data loss. However, as an user my probability is still the same | 15:02 |
| cschwede | notmyname: i was somewhat worried when i saw that on HN, took me a while to spot the difference | 15:03 |
| notmyname | oh yeah? I just woke up and read it :-) | 15:03 |
| tdasilva | good morning | 15:03 |
| cschwede | notmyname: if you have a VERY LARGE cluster, you would expect a data loss probability of close to 1, right? it's just the law of the big numbers IIRC | 15:04 |
| notmyname | oh so kinda like the birthday paradox. more people increases the chance that *some* pair has the same birthday, but there's always still the 1/365 chance that someone has the same birthday as *you* | 15:04 |
| cschwede | but the probability to loose one of _my_ objects is still like 1*10-9 or whatever my number is | 15:05 |
| *** bkopilov has quit IRC | 15:05 | |
| notmyname | that's not very comforting, though. | 15:07 |
| notmyname | because we aren't supposed to protect for a given partition. we're supposed to protect for any data loss for all partitions | 15:09 |
| *** bkopilov has joined #openstack-swift | 15:11 | |
| tdasilva | mm..interesting, i guess like the author i'd be also interested in hearing what's happening in practice? | 15:13 |
| *** sams-gleb has quit IRC | 15:16 | |
| *** caiobrentano has joined #openstack-swift | 15:16 | |
| *** sams-gleb has joined #openstack-swift | 15:17 | |
| * notmyname needs to continue to get ready for the day | 15:17 | |
| *** mvk has joined #openstack-swift | 15:18 | |
| *** vinsh has joined #openstack-swift | 15:20 | |
| *** sams-gleb has quit IRC | 15:21 | |
| *** sams-gleb has joined #openstack-swift | 15:33 | |
| *** mvk has quit IRC | 15:40 | |
| jordanP | the comments on that article are informative | 15:49 |
| *** mvk has joined #openstack-swift | 15:53 | |
| *** acoles_ is now known as acoles | 15:54 | |
| *** SkyRocknRoll has quit IRC | 16:22 | |
| *** klrmn has joined #openstack-swift | 16:22 | |
| *** rcernin has quit IRC | 16:24 | |
| *** ppai has quit IRC | 16:28 | |
| *** tesseract has quit IRC | 16:58 | |
| *** garyj has joined #openstack-swift | 17:02 | |
| notmyname | good morning (for realz) | 17:05 |
| *** jordanP has quit IRC | 17:16 | |
| *** chsc has joined #openstack-swift | 17:17 | |
| *** chsc has joined #openstack-swift | 17:17 | |
| *** klrmn has quit IRC | 17:18 | |
| *** mvk has quit IRC | 17:24 | |
| *** StraubTW has joined #openstack-swift | 17:37 | |
| *** garyj has quit IRC | 17:40 | |
| *** JimCheung has joined #openstack-swift | 17:41 | |
| *** rledisez has quit IRC | 17:45 | |
| *** garyj has joined #openstack-swift | 17:48 | |
| *** newmember has quit IRC | 17:54 | |
| jrichli | notmyname: thanks for pointing out the OS statement, and new priority reviews list. bkeller` and I are at San Jose with clu and tqtran for the next few days. | 18:04 |
| timburke | good morning | 18:04 |
| jrichli | timburke: good morning | 18:05 |
| acoles | notmyname: feeling better? | 18:06 |
| *** mvk has joined #openstack-swift | 18:07 | |
| timburke | notmyname: that sounds more like keystone's discovery api -- i think he might want to look at https://github.com/openstack/keystoneauth/blob/master/keystoneauth1/discover.py ? | 18:09 |
| timburke | iirc, the generic password plugin should already handle the v2/v3 split fairly seemlessly | 18:11 |
| *** klrmn has joined #openstack-swift | 18:12 | |
| notmyname | acoles: I am feeling much more normal. thank | 18:12 |
| notmyname | *thanks | 18:13 |
| acoles | notmyname: good to hear | 18:13 |
| acoles | clayg: ping | 18:14 |
| *** htruta` is now known as htruta | 18:23 | |
| clayg | acoles: pong | 18:23 |
| acoles | clayg: hi. I was just typing up some more comment on patch https://review.openstack.org/#/c/419787 | 18:24 |
| patchbot | patch 419787 - swift - Better optimistic lock in get_hashes | 18:24 |
| clayg | acoles: I played with a follup patch over the weekened - trying to figure out a better way to represent "None"/all-hashes-are-invalid *on-disk* | 18:24 |
| acoles | clayg: yes, that! just left a link to a gist, I wonder how similar our solutions are? | 18:24 |
| clayg | but I'm having a hard time dreaming up a test where {'updated': now()} is acctually wrong - part of the "problem" seems to be the way that get_hashes passes in the do_listdir when it has to recurse because it looses the optimistic lock. | 18:25 |
| clayg | acoles: my patch wrote {'updated': now(), 'valid': False} - what about yours? I didn't really *like* mine. | 18:26 |
| acoles | clayg: hehe, at some point today i wrote exactly that! | 18:26 |
| clayg | my first attempt serialized it into a dict - but revivified as a SuffixHashes object | 18:26 |
| clayg | hated that too | 18:26 |
| acoles | clayg: I ended up with 'created': time.time() when writing a hashes.pkl that does not have results of suffix calcs (i.e. a "fresh" hashes.pkl) | 18:27 |
| clayg | acoles: nice - sounds like you might have gotten further than I did! Do you love it!?!? | 18:27 |
| acoles | clayg: we have to (a) mark a hashes.pkl that is fresh and (b) do it differently in every process that writes a fresh pkl | 18:27 |
| clayg | oh you have a test - you definately got further | 18:27 |
| *** zaitcev has joined #openstack-swift | 18:28 | |
| *** ChanServ sets mode: +v zaitcev | 18:28 | |
| clayg | acoles: essentially it seems yes | 18:28 |
| acoles | clayg: I try to stay dispassionate about my code ;) | 18:28 |
| acoles | too many htings i have loved have been torn down by greater minds ;) | 18:28 |
| clayg | acoles: I hate 99% of my code - so the 1% that is not obvious crap I tend to get pretty excited about. | 18:28 |
| *** tqtran has joined #openstack-swift | 18:29 | |
| acoles | lol | 18:29 |
| *** vills_ has quit IRC | 18:30 | |
| acoles | clayg: what I do like is that once the follow up patch lands, consolidate_hashes either writes a pkl and returns a dict, no more None. | 18:30 |
| acoles | or raises exception | 18:31 |
| acoles | (with the gist that is) | 18:31 |
| clayg | acoles: i'm still going over the comments - should I jump ahead? | 18:33 |
| acoles | clayg: so when I had just 'valid': False, IIRC the follow up patch https://review.openstack.org/#/c/426336/3 test change would then fail | 18:33 |
| patchbot | patch 426336 - swift - Fix race when consolidating new partition | 18:33 |
| acoles | clayg: which drove me to created=time.time(), but with care about when that gets set - refresh page and see my last comment on gerrit | 18:36 |
| acoles | clayg: I'm not sure if I picked up all Pavel's and mahatic comments in my gist, so there may be other comments to pay attention to, but the gist is just about all I have to offer today | 18:37 |
| clayg | acoles: ok, the test looks like something. I didn't immediately fall in love with created key - i see a lot of handling for "if created in hashes" and "pop('created', None)" - but the test is a huge head start | 18:39 |
| clayg | One of things I hated about my patch is where intialization was just "hashes = None" before I changed it to {'valid': False'} | 18:40 |
| acoles | clayg: I tried and failed to overload 'updated' key - see my gerrit comments | 18:40 |
| clayg | acoles: I can definately vouche overloading 'updated' is a path to madness | 18:41 |
| clayg | acoles: ok, tactics | 18:41 |
| clayg | acoles: if I manage to come up with something I like today - should I push over patch 419787? | 18:42 |
| patchbot | https://review.openstack.org/#/c/419787/ - swift - Better optimistic lock in get_hashes | 18:42 |
| acoles | clayg: yes. I am dome for today | 18:42 |
| acoles | done* | 18:42 |
| clayg | acoles: does the test at https://gist.github.com/alistairncoles/648c0d8e7f35f1bc4d5cf8994d8a9ce0#file-gist-diff-L120 demonstrate a *regression* or just another failure? | 18:42 |
| openstackgerrit | Tim Burke proposed openstack/swift: Warn about using EC with isa_l_rs_vand and nparity >= 5 https://review.openstack.org/425496 | 18:43 |
| clayg | acoles: what's the status on patch 426336 - that test works on master - but sadly only because of the double-rehash-everytime bug | 18:43 |
| patchbot | https://review.openstack.org/#/c/426336/ - swift - Fix race when consolidating new partition | 18:43 |
| clayg | we have to make *some* kind of progress in this method at some point - i'm worried great is the enemy of good-enough? (obviously I'm tired and have other patches I'm worried about too) | 18:44 |
| clayg | if we have a clear path of what's left to fix and we think we manage the diff - i'm all for it - but I'm slowly loosing track of all the different problems we're fixing in this patch | 18:44 |
| clayg | got more races than a day at the tracks | 18:44 |
| acoles | clayg: the test in the gist passes on master (small sample) | 18:46 |
| *** foexle has quit IRC | 18:46 | |
| acoles | clayg: patch 426336 I think can stay as a follow on, to keep things simple. | 18:47 |
| patchbot | https://review.openstack.org/#/c/426336/ - swift - Fix race when consolidating new partition | 18:47 |
| clayg | acoles: ok, wfm | 18:47 |
| clayg | I don't have anything about patch 334719 loaded into my head and no bandwidth to load it today I don't think | 18:49 |
| patchbot | https://review.openstack.org/#/c/334719/ - swift - Preserve X-Static-Large-Object from .data file aft... | 18:49 |
| clayg | there's also a follow up patch apparently that needs to be squashed - if anyone knows what's going on there and wants to push over it'd be greatly appreciated | 18:49 |
| clayg | if not - I'll try not to forget about it again | 18:49 |
| acoles | clayg: the change you may in diskfile on patch 426336 was exactly what I had in mind but before latest rev of patch 419787 it caused some errors, but now it looks good because we're always writing a hashes.pkl | 18:49 |
| patchbot | https://review.openstack.org/#/c/426336/ - swift - Fix race when consolidating new partition | 18:49 |
| patchbot | https://review.openstack.org/#/c/419787/ - swift - Better optimistic lock in get_hashes | 18:50 |
| clayg | yeah, np | 18:50 |
| acoles | clayg: I need to go, sure youll do the right thing | 18:51 |
| *** acoles is now known as acoles_ | 18:53 | |
| *** geaaru has quit IRC | 19:27 | |
| timburke | notmyname: fyi https://review.openstack.org/426894 | 19:35 |
| patchbot | patch 426894 - releases - stable/ocata branch for python-swiftclient | 19:35 |
| timburke | notmyname: oh yeah, did we ever get updates on the project mascot? is it actually a swift now? | 19:38 |
| notmyname | no word yet on mascot | 19:40 |
| notmyname | timburke: oh, right. we've got to specifically ask for that stable release now, right? | 19:40 |
| notmyname | thanks | 19:40 |
| timburke | yup | 19:41 |
| notmyname | ok, I'll add my + vote | 19:41 |
| *** oshritf__ has quit IRC | 19:41 | |
| * notmyname goes to lunch | 19:42 | |
| openstackgerrit | Tim Burke proposed openstack/swift: Warn about using EC with isa_l_rs_vand and nparity >= 5 https://review.openstack.org/425496 | 19:58 |
| * timburke grumbles | 19:58 | |
| *** silor has quit IRC | 20:03 | |
| *** ChubYann has joined #openstack-swift | 20:06 | |
| *** garyj has quit IRC | 20:21 | |
| *** vills has joined #openstack-swift | 20:26 | |
| *** vills has quit IRC | 20:26 | |
| *** newmember has joined #openstack-swift | 20:30 | |
| *** NM has quit IRC | 20:34 | |
| notmyname | timburke: looks like the branch already landed | 20:36 |
| timburke | yup. so we can go land https://review.openstack.org/#/c/426902/ | 20:40 |
| patchbot | patch 426902 - python-swiftclient (stable/ocata) - Update .gitreview for stable/ocata | 20:40 |
| notmyname | timburke: cool. thanks for that too | 20:41 |
| timburke | notmyname: well now it's bots all the way down | 20:41 |
| mattoliverau | Morning | 21:05 |
| timburke | clayg: i'd feel a whole lot better about https://review.openstack.org/#/c/425441/ with something like http://paste.openstack.org/show/596937/ applied... | 21:12 |
| patchbot | patch 425441 - swift - Do not revert fragments to handoffs | 21:12 |
| timburke | as it is, i'm not sure i know enough about the reconstructor to say definitively that it's a good idea | 21:12 |
| *** vint_bra has quit IRC | 21:23 | |
| clayg | timburke: I think that mostly ends up moving it twice in practice - but possibly get it into the handoff chain sooner | 21:30 |
| clayg | might be attractive if you run a really low m | 21:30 |
| clayg | definately better than letting them flop around forever! | 21:30 |
| timburke | clayg: again, it just seems like we're hurting availability without something like that. might want some constant (0 < C < 1?) in front of that replica_count, though? | 21:32 |
| *** NM has joined #openstack-swift | 21:33 | |
| clayg | i think unless the primary disk is unmounted the absolute best thing to do is wait | 21:34 |
| *** Jeffrey4l__ has joined #openstack-swift | 21:34 | |
| *** Jeffrey4l_ has quit IRC | 21:35 | |
| clayg | after a rebalance you're used to an "availability issue" - the best was to avoid a *durability* issue is to get from "rebalancing" to "rebalanced" asap - that means only moving them once | 21:35 |
| *** NM has quit IRC | 21:36 | |
| clayg | IME - some additional experimentation may be in order - but i'm sure the patch works better than master - maybe that's not good enough? | 21:36 |
| *** catintheroof has quit IRC | 21:40 | |
| *** catintheroof has joined #openstack-swift | 21:42 | |
| *** NM has joined #openstack-swift | 21:42 | |
| *** newmember has quit IRC | 21:48 | |
| timburke | clayg: when we were talking about test_get_more_nodes, you said that we wanted to make sure the handoff iter stayed relatively stable through rebalances, particularly at the start. do we have any sort of feel for the probabilities involved here? | 21:49 |
| timburke | seems like if the handoff list is probably mostly the same, then (1) you probably *won't* have an availability issue, (2) the reconstructor would most likely hit the i'm-a-handoff path and wait, and (3) having data that's not fairly high up in priority is rather bad | 21:49 |
| *** NM has quit IRC | 21:52 | |
| timburke | (...data *on a node* that's not fairly high up in priority...) | 21:52 |
| *** catinthe_ has joined #openstack-swift | 21:52 | |
| *** catintheroof has quit IRC | 21:53 | |
| clayg | timburke: I think maybe it depends on how many disks you have (or are adding?) - I don't think it's particuarlly common (although maybe not exceptionally rare depending on the # of disk) that an old primary becomes a handoff (all disks are handoffs at some point, it's just a matter of depth) | 21:54 |
| clayg | timburke: anyway - i'm not sure I follow the significance of that line of reasoning - I don't think I'm strongly against moving parts "between once and twice" and rebalance - it's close enough to "once" from "betweence once and infinity" that it's probably good enough for me! | 22:00 |
| *** sams-gleb has quit IRC | 22:01 | |
| timburke | blerg, right, the majority of the time it'll be an old-primary rather than an old-handoff-still-within-request_node_count (see why i shorten that to handoff? maybe we need a separate term there...) | 22:02 |
| timburke | ok, i think i'm getting more on board with it as-is | 22:03 |
| *** catintheroof has joined #openstack-swift | 22:04 | |
| *** catinthe_ has quit IRC | 22:07 | |
| *** foexle has joined #openstack-swift | 22:09 | |
| clayg | timburke: I think the replicator does this cute trick where it has like "attempts" - then it += 1 when it sees a 507 response | 22:13 |
| clayg | idk, i'm still working on SuffixHashes - I'll keep thinking about when/how to revert to handoff | 22:15 |
| clayg | timburke: thanks! | 22:15 |
| timburke | hmmm... seems like "old primary becomes a handoff" would be a rather nice thing to have... | 22:15 |
| *** vint_bra has joined #openstack-swift | 22:17 | |
| *** catinthe_ has joined #openstack-swift | 22:17 | |
| clayg | I've had any good thoughts on how to do it - it's more attractive on GET than PUT | 22:18 |
| *** catinth__ has joined #openstack-swift | 22:18 | |
| *** catintheroof has quit IRC | 22:18 | |
| timburke | idk, seems like on PUT it'd still be nice to have that handoff get rid of any overwritten data quickly | 22:19 |
| timburke | clayg: i think i'm on board, though. want me to just go change the bug to 1653169 and call it done? is it worth cleaning up those unnecessary breaks? | 22:19 |
| timburke | we never create revert jobs on primaries, right? that seems like it'd be a silly thing to do | 22:20 |
| *** catinthe_ has quit IRC | 22:22 | |
| clayg | timburke: I think you raised some good points in comments - i'd like to give them some more thought - I can push something up by EOD - do you think you'd have time to give it another round tmrw? | 22:22 |
| timburke | yeah, most likely. but if i'm just fixing the bug #, we keep cschwede's +2 and we can drop those breaks (turns out neither check is necessary) later | 22:26 |
| clayg | timburke: sorry - my heads not really in game - i trust your judgement | 22:27 |
| *** foexle has quit IRC | 22:27 | |
| *** catinth__ has quit IRC | 22:29 | |
| timburke | clayg: no worries; i know you've got hashes on the mind | 22:29 |
| *** sams-gleb has joined #openstack-swift | 22:42 | |
| *** sams-gleb has quit IRC | 22:42 | |
| *** sams-gleb has joined #openstack-swift | 22:42 | |
| *** sams-gleb has quit IRC | 22:43 | |
| timburke | i take back my remark on silliness -- it's a perfectly reasonable thing to do if the primary has multiple frags, in which case i don't like https://github.com/openstack/swift/blob/2.12.0/swift/obj/reconstructor.py#L659-L661 ... | 22:49 |
| timburke | i think i need to go read more about the reconstructor | 22:49 |
| timburke | looking at http://docs.openstack.org/developer/swift/overview_erasure_code.html -- anyone know what's meant by "Additionally, its [sic] not always the case that the processing of a particular suffix directory means one or the other for the entire directory"? one or the other *what*? | 23:06 |
| *** StraubTW has quit IRC | 23:06 | |
| clayg | one or the other "job types" - replicator has update and updated_deleted - each part goes into once - reconstructor can have a part generate jobs of both types | 23:18 |
| clayg | I think that was all it was getting at | 23:18 |
| clayg | ^ from the last paragraph in http://docs.openstack.org/developer/swift/overview_erasure_code.html#the-reconstructor | 23:19 |
| clayg | honestly I totally forgot that document went into that detail - some of those reconstructor changes could be introducing doc regressions - i should audit... | 23:20 |
| *** kei_yama has joined #openstack-swift | 23:30 | |
| openstackgerrit | Tim Burke proposed openstack/swift: Clean up EC overview docs a bit https://review.openstack.org/426971 | 23:30 |
| *** sileht has quit IRC | 23:37 | |
| *** chsc has quit IRC | 23:37 | |
| *** sileht has joined #openstack-swift | 23:41 | |
| *** sams-gleb has joined #openstack-swift | 23:43 | |
| *** sams-gleb has quit IRC | 23:48 | |
| *** catintheroof has joined #openstack-swift | 23:50 | |
| *** jamielennox is now known as jamielennox|away | 23:58 | |
| *** jamielennox|away is now known as jamielennox | 23:59 | |
Generated by irclog2html.py 2.14.0 by Marius Gedminas - find it at mg.pov.lt!