| *** jamielennox|away is now known as jamielennox | 00:05 | |
| *** klrmn has joined #openstack-swift | 00:11 | |
| *** dmorita has quit IRC | 01:00 | |
| *** vint_bra has quit IRC | 01:20 | |
| *** dmorita has joined #openstack-swift | 01:26 | |
| *** chsc has joined #openstack-swift | 01:29 | |
| *** dmorita has quit IRC | 01:30 | |
| *** chsc has quit IRC | 01:41 | |
| *** mingyu has joined #openstack-swift | 01:49 | |
| *** dmorita has joined #openstack-swift | 02:16 | |
| *** cebreidian has quit IRC | 02:21 | |
| *** dmorita has quit IRC | 02:21 | |
| *** dmorita has joined #openstack-swift | 02:27 | |
| *** cebreidian has joined #openstack-swift | 02:30 | |
| *** mingyu has quit IRC | 02:31 | |
| *** dmorita has quit IRC | 02:32 | |
| *** mingyu has joined #openstack-swift | 02:46 | |
| *** jerrygb has joined #openstack-swift | 02:55 | |
| *** dmorita has joined #openstack-swift | 03:29 | |
| *** dmorita has quit IRC | 03:33 | |
| *** early has joined #openstack-swift | 03:35 | |
| *** early` has quit IRC | 03:36 | |
| *** cebreidian has quit IRC | 03:42 | |
| *** csmart has quit IRC | 03:43 | |
| *** jerrygb has quit IRC | 04:03 | |
| *** csmart has joined #openstack-swift | 04:17 | |
| *** psachin has joined #openstack-swift | 04:24 | |
| *** mingyu has quit IRC | 04:43 | |
| *** cebreidian has joined #openstack-swift | 04:45 | |
| *** chsc has joined #openstack-swift | 04:49 | |
| *** chsc has quit IRC | 04:54 | |
| *** cebreidian has quit IRC | 05:00 | |
| *** cebreidian has joined #openstack-swift | 05:09 | |
| openstackgerrit | Matthew Oliver proposed openstack/swift: Add container sharding to Swift containers https://review.openstack.org/423906 | 05:09 |
|---|---|---|
| mattoliverau | ^^ very rough WIP that may not actaully run ATM. It's a rebase of the HEAD of the current POC. Tests, documentation, debugging still required.. but thought it's time to put it somethere. Will go through and leave some comments in the code as required. but I'm thinking reviews should start once I've finished writing up the dev docs that will go with the patch... so everyone knows how it works.. otherwise it'll be a | 05:12 |
| mattoliverau | little daunting. And then.. please go fix my dodgy code that has 5 POCs cruft probably thoughout it. | 05:12 |
| *** cebreidian has quit IRC | 05:14 | |
| jrichli | mattoliverau: woohoo! Sharding patch \o/ ... I need to watch your talk from lca. I guess they are available now? I didnt get to watch the live stream. | 05:19 |
| mattoliverau | jrichli: they are. They should be on youtube | 05:19 |
| jrichli | great. i loved the last sharding one - great explanations of the phases you had gone through | 05:20 |
| mattoliverau | jrichli: \o/.. still a quite aways to go.. but not sitting as POCs anymore, so finally in gerrit | 05:20 |
| mattoliverau | This one went better, ie I spoke and the presentation was better (in my opinion) Tho the introduction (introducing me) was crap.. I was kind dumbfounding by that for the first minute or so. | 05:21 |
| mattoliverau | instead of introducing the talk by reading the title, they said "some rackspace thing" which isn't totally what it isn't. | 05:22 |
| jrichli | well i am excited to hear how things ended up being done - and then i can understand enough to look at the patch, i hope :-) | 05:23 |
| mattoliverau | jrichli: you mean fix my code to make it better right ;) | 05:23 |
| mattoliverau | ? | 05:23 |
| jrichli | ah ... yeah. ;-) I am just gonna focus on understanding right now! | 05:24 |
| mattoliverau | lol | 05:24 |
| *** klrmn has quit IRC | 05:24 | |
| *** cebreidian has joined #openstack-swift | 05:31 | |
| openstackgerrit | Tone Zhang proposed openstack/swift: IO Priority support on the aarch64 architecture https://review.openstack.org/423783 | 05:37 |
| *** zaitcev has quit IRC | 05:37 | |
| *** ppai has joined #openstack-swift | 05:42 | |
| *** jerrygb has joined #openstack-swift | 06:04 | |
| *** jerrygb has quit IRC | 06:08 | |
| *** bkopilov has joined #openstack-swift | 06:23 | |
| *** hseipp has joined #openstack-swift | 07:07 | |
| *** hseipp has quit IRC | 07:16 | |
| *** ChubYann has quit IRC | 07:18 | |
| *** McMurlock1 has joined #openstack-swift | 07:19 | |
| *** McMurlock1 has quit IRC | 07:25 | |
| *** tesseract has joined #openstack-swift | 07:31 | |
| *** sams-gleb has joined #openstack-swift | 07:38 | |
| *** oshritf has joined #openstack-swift | 07:52 | |
| *** rledisez has joined #openstack-swift | 08:15 | |
| *** hseipp has joined #openstack-swift | 08:15 | |
| *** pcaruana has joined #openstack-swift | 08:17 | |
| *** ppai has quit IRC | 08:29 | |
| *** geaaru has joined #openstack-swift | 08:29 | |
| *** hseipp has quit IRC | 08:51 | |
| openstackgerrit | Christian Schwede proposed openstack/swift: Add support to increase object ring partition power https://review.openstack.org/337297 | 08:55 |
| *** abhitechie has joined #openstack-swift | 09:10 | |
| *** abhinavtechie has joined #openstack-swift | 09:13 | |
| *** abhitechie has quit IRC | 09:13 | |
| *** abhinavtechie has quit IRC | 09:15 | |
| *** abhinavtechie has joined #openstack-swift | 09:15 | |
| *** hseipp has joined #openstack-swift | 09:16 | |
| *** cbartz has joined #openstack-swift | 09:19 | |
| *** yarkot has quit IRC | 09:29 | |
| *** mingyu has joined #openstack-swift | 09:37 | |
| *** abhinavtechie has quit IRC | 09:45 | |
| *** abhitechie has joined #openstack-swift | 09:46 | |
| *** abhitechie has quit IRC | 10:03 | |
| *** abhitechie has joined #openstack-swift | 10:03 | |
| *** acoles_ is now known as acoles | 10:07 | |
| *** mvk has quit IRC | 10:12 | |
| *** abhinavtechie has joined #openstack-swift | 10:17 | |
| *** abhitechie has quit IRC | 10:17 | |
| *** mariusv has quit IRC | 10:19 | |
| *** mariusv has joined #openstack-swift | 10:20 | |
| *** mariusv has quit IRC | 10:20 | |
| *** mariusv has joined #openstack-swift | 10:20 | |
| *** sams-gleb has quit IRC | 10:25 | |
| *** sams-gleb has joined #openstack-swift | 10:26 | |
| *** sams-gleb has quit IRC | 10:30 | |
| *** zhugaoxiao has quit IRC | 10:34 | |
| *** sams-gleb has joined #openstack-swift | 10:45 | |
| *** mvk has joined #openstack-swift | 10:46 | |
| *** vint_bra has joined #openstack-swift | 11:15 | |
| *** vint_bra has quit IRC | 11:16 | |
| *** dmorita has joined #openstack-swift | 11:18 | |
| *** ganders has joined #openstack-swift | 11:21 | |
| *** dmorita has quit IRC | 11:22 | |
| *** caiobrentano_ has joined #openstack-swift | 11:36 | |
| *** abhinavtechie has quit IRC | 11:38 | |
| *** vint_bra has joined #openstack-swift | 12:13 | |
| *** mingyu has quit IRC | 12:15 | |
| *** mingyu has joined #openstack-swift | 12:20 | |
| *** kei_yama has quit IRC | 12:33 | |
| *** mingyu has quit IRC | 12:39 | |
| *** vint_bra has quit IRC | 12:41 | |
| *** catintheroof has joined #openstack-swift | 12:45 | |
| *** vint_bra has joined #openstack-swift | 12:46 | |
| *** oshritf has quit IRC | 12:48 | |
| *** catintheroof has quit IRC | 12:50 | |
| *** psachin has quit IRC | 12:50 | |
| *** catintheroof has joined #openstack-swift | 12:58 | |
| *** hseipp has quit IRC | 13:05 | |
| *** vint_bra has quit IRC | 13:24 | |
| *** oshritf has joined #openstack-swift | 13:24 | |
| *** vint_bra has joined #openstack-swift | 13:24 | |
| *** oshritf has quit IRC | 13:34 | |
| *** caiobrentano_ has quit IRC | 13:34 | |
| *** caiobrentano_ has joined #openstack-swift | 13:35 | |
| *** jordanP has joined #openstack-swift | 13:38 | |
| *** mingyu has joined #openstack-swift | 13:39 | |
| *** oshritf has joined #openstack-swift | 13:39 | |
| *** mingyu has quit IRC | 13:44 | |
| *** yangzb09 has joined #openstack-swift | 13:55 | |
| tdasilva | good morning | 14:11 |
| *** jerrygb has joined #openstack-swift | 14:14 | |
| *** jerrygb_ has joined #openstack-swift | 14:15 | |
| *** jerrygb__ has joined #openstack-swift | 14:18 | |
| *** jerrygb has quit IRC | 14:19 | |
| *** jerrygb_ has quit IRC | 14:20 | |
| *** catinthe_ has joined #openstack-swift | 14:29 | |
| *** catintheroof has quit IRC | 14:32 | |
| *** mingyu has joined #openstack-swift | 14:40 | |
| *** hseipp has joined #openstack-swift | 14:43 | |
| *** mingyu has quit IRC | 14:45 | |
| *** vinsh has quit IRC | 14:46 | |
| *** mingyu has joined #openstack-swift | 14:46 | |
| *** vinsh has joined #openstack-swift | 14:47 | |
| *** sams-gleb has quit IRC | 15:00 | |
| *** sams-gleb has joined #openstack-swift | 15:01 | |
| *** jordanP has quit IRC | 15:03 | |
| *** jordanP has joined #openstack-swift | 15:03 | |
| *** sams-gleb has quit IRC | 15:05 | |
| *** oshritf has quit IRC | 15:15 | |
| *** sams-gleb has joined #openstack-swift | 15:17 | |
| *** oshritf has joined #openstack-swift | 15:18 | |
| *** oshritf has quit IRC | 15:19 | |
| *** vinsh_ has joined #openstack-swift | 15:21 | |
| openstackgerrit | Merged openstack/swift: Simplify get_different_suffix_df args https://review.openstack.org/423501 | 15:21 |
| *** vinsh has quit IRC | 15:24 | |
| *** oshritf has joined #openstack-swift | 15:26 | |
| *** mingyu has quit IRC | 15:29 | |
| *** mingyu has joined #openstack-swift | 15:29 | |
| *** _JZ_ has joined #openstack-swift | 15:32 | |
| *** oshritf_ has joined #openstack-swift | 15:35 | |
| *** oshritf has quit IRC | 15:37 | |
| *** jerrygb has joined #openstack-swift | 15:45 | |
| *** mingyu has quit IRC | 15:46 | |
| *** jerrygb__ has quit IRC | 15:47 | |
| *** oshritf_ has quit IRC | 15:49 | |
| acoles | tdasilva: I pushed a change over patch 423508, hope that's ok | 16:02 |
| patchbot | https://review.openstack.org/#/c/423508/ - openstack-infra/project-config - Update swift functional job list | 16:02 |
| tdasilva | acoles: thank you, i saw that earlier, great catch! | 16:02 |
| *** Jeffrey4l_ has quit IRC | 16:06 | |
| *** Jeffrey4l_ has joined #openstack-swift | 16:06 | |
| *** sanchitmalhotra1 has joined #openstack-swift | 16:07 | |
| *** sanchitmalhotra has quit IRC | 16:08 | |
| *** sanchitmalhotra1 is now known as sanchitmalhotra | 16:08 | |
| openstackgerrit | Alistair Coles proposed openstack/swift: Fix race in new partitions detecting new/invalid suffixes. https://review.openstack.org/418690 | 16:09 |
| acoles | clayg: kota_ ^^ rebased, fixed merge conflicts, probably just needs clayg's blessing on the test changes | 16:10 |
| *** pcaruana has quit IRC | 16:13 | |
| *** zaitcev has joined #openstack-swift | 16:15 | |
| *** ChanServ sets mode: +v zaitcev | 16:15 | |
| *** mvk has quit IRC | 16:18 | |
| *** klrmn has joined #openstack-swift | 16:20 | |
| *** hoonetorg has joined #openstack-swift | 16:42 | |
| notmyname | good morning | 16:44 |
| *** hoonetorg has quit IRC | 16:45 | |
| tdasilva | notmyname: o/ | 16:46 |
| *** mingyu has joined #openstack-swift | 16:46 | |
| *** hoonetorg has joined #openstack-swift | 16:47 | |
| *** chsc has joined #openstack-swift | 16:50 | |
| *** chsc has joined #openstack-swift | 16:50 | |
| *** vint_bra has quit IRC | 16:51 | |
| *** caiobren_ has joined #openstack-swift | 16:51 | |
| *** mingyu has quit IRC | 16:52 | |
| *** caiobre__ has joined #openstack-swift | 16:54 | |
| *** caiobrentano_ has quit IRC | 16:54 | |
| *** caiobren_ has quit IRC | 16:55 | |
| *** david-lyle_ has joined #openstack-swift | 16:57 | |
| *** hseipp has quit IRC | 16:57 | |
| *** chsc has quit IRC | 16:57 | |
| *** cbartz has left #openstack-swift | 16:59 | |
| *** yangzb09 has quit IRC | 16:59 | |
| *** caiobre__ has quit IRC | 17:03 | |
| *** david-lyle_ has quit IRC | 17:03 | |
| *** vint_bra has joined #openstack-swift | 17:05 | |
| *** caiobrentano_ has joined #openstack-swift | 17:05 | |
| *** vint_bra has quit IRC | 17:10 | |
| *** vint_bra has joined #openstack-swift | 17:10 | |
| *** tesseract has quit IRC | 17:15 | |
| *** klrmn has quit IRC | 17:19 | |
| *** caiobrentano_ has quit IRC | 17:22 | |
| *** caiobrentano_ has joined #openstack-swift | 17:23 | |
| clayg | good morning! | 17:24 |
| clayg | acoles: oh, i was thinking I had a bunch of rebasing to do this am | 17:24 |
| clayg | acoles: I looked at some of the test changes - it wasn't clear why all of them were needed (how they made the tests better) - but if you like them better; I'm sure it's fine! | 17:25 |
| notmyname | mattoliverau: thanks for pushing up https://review.openstack.org/#/c/423906/ | 17:25 |
| patchbot | patch 423906 - swift - Add container sharding to Swift containers | 17:25 |
| clayg | acoles: you're always saying the mountains of tests are what makes it so easy to maintain swift and more benign implementation details they make assertions about the easier it is to refactor the code behind the interfaces! | 17:26 |
| clayg | ah... i *am* feeling better! | 17:26 |
| *** mvk has joined #openstack-swift | 17:26 | |
| clayg | notmyname: have you looked at it!? | 17:26 |
| clayg | notmyname: i'm sort of scared too :'( | 17:27 |
| clayg | notmyname: maybe global EC or symlinks first? | 17:27 |
| clayg | notmyname: or part power - but mattoliverau's been working on that too! | 17:27 |
| notmyname | clayg: I haven't looked at the patch yet (I just saw it), but I talked to mattoliverau about it last week. from what he's said, I think the idea's pretty good and I'm glad we can start looking at actual code | 17:29 |
| clayg | notmyname: yeah! that's the spirit! | 17:29 |
| notmyname | however, I personally don't feel too much pressure to sit down with it in great detail before the PTG (although having a baseline for discussion at the PTG will be good) | 17:30 |
| acoles | clayg: the ones that I remember bothering me were like https://review.openstack.org/#/c/418690/5/test/unit/obj/test_auditor.py@859 | 17:30 |
| patchbot | patch 418690 - swift - Fix race in new partitions detecting new/invalid s... | 17:30 |
| clayg | yeah, the auditor - that was the bulk of the change - what *was* going on there!? | 17:30 |
| notmyname | like you said, some of the other stuff that's going on is much closer to being done and should take priority. global EC, part power increase, and maybe symlinks | 17:30 |
| *** JimCheung has joined #openstack-swift | 17:31 | |
| acoles | was this the patch where I actually said "maybe clayg was deliberately avoiding digging around under the hood too much with assertions" ? or words to that effect :o | 17:31 |
| clayg | well, no i see your point - if the existing test was coupled with an assumption about the behavior of invalidate_hash (which I changed) | 17:32 |
| clayg | it *is* sorta bs for the tests to keep living on like that - good catch | 17:32 |
| clayg | so what did you come up with for the fix? assertEqual(mock_invalidate_hash.call_arg_list, []) ???? | 17:33 |
| *** chsc has joined #openstack-swift | 17:33 | |
| *** chsc has joined #openstack-swift | 17:33 | |
| acoles | TBH in the auditor tests they should probably have been mocking invalidate_hash on master since the goal is to verify that invalidate_hash is not called | 17:33 |
| acoles | um, no I think I stuck with the pattern of verifying that the state of hashes.invalid is how you'd expect it, which is of course as you say more brittle to implementation changes :/ | 17:35 |
| acoles | but no worse than what was on master (is that a valid defense? ;) ) | 17:36 |
| clayg | ok, w/e - i'll do a follow up to change it to the mocks! | 17:36 |
| clayg | nice work! | 17:36 |
| acoles | clayg: looks like bulk of the test changes I made was from patchset 5 to 6 | 17:39 |
| clayg | cool - i'm just going to make sure the cleanup looks legit and then +A and move onto rebasing all the other stuff | 17:40 |
| clayg | acoles: there was one patch I saw where you started to "standardize handling of invalid/non-existing hashes/pkl" | 17:41 |
| clayg | acoles: i'm my fever induced visions I also came to that conclusion (assuming I was reading the patch correctly!) | 17:41 |
| acoles | clayg: this one maybe patch 422076 | 17:42 |
| patchbot | https://review.openstack.org/#/c/422076/ - swift - Handle unreadable hashes.pkl same way as non-exist... | 17:42 |
| clayg | acoles: I'm also became quite bummed out about except Exception around consolidate_hashes (which I know now has strong tests :P) | 17:42 |
| clayg | acoles: I don't think it was supposed to be there | 17:42 |
| acoles | yeah there is still some cruft in there | 17:42 |
| clayg | acoles: I think the original pre consolidate hashes code had more reasonable exception handling | 17:42 |
| clayg | particuarlly - the lockdir call (which used to be around reading hashes, and is now around consoidate hashes) used to be able to raise the timeout all the way out to the REPLICATE call | 17:43 |
| clayg | now it carries on and does the rehash (making more io contention in the dir) - THEN it will raise the lockdir when it finally tries to writeback | 17:43 |
| acoles | clayg: so my patch 422076 does need a rebase but I didn't think any other of yours did, unless gerrit hasn't caught up with them yet | 17:44 |
| patchbot | https://review.openstack.org/#/c/422076/ - swift - Handle unreadable hashes.pkl same way as non-exist... | 17:44 |
| clayg | getting git of that pokemon exception handling there will make it so we can safely put the getinode call under the lockdir! | 17:44 |
| acoles | you weren't really sick were you, just plotting :) | 17:44 |
| clayg | acoles: i think it's more like you said - the sickness came out of worry over get_hashes | 17:46 |
| clayg | acoles: ok, well I don't even have all the patches up yet | 17:46 |
| acoles | lol, a bad case of hashitis | 17:46 |
| clayg | acoles: i'm going to just work on merging patch 418690 | 17:46 |
| patchbot | https://review.openstack.org/#/c/418690/ - swift - Fix race in new partitions detecting new/invalid s... | 17:47 |
| acoles | clayg: +1 to that | 17:47 |
| clayg | acoles: then i'm mostly thinking about patch 419787 | 17:47 |
| patchbot | https://review.openstack.org/#/c/419787/ - swift - Better optimistic lock in get_hashes | 17:47 |
| clayg | which I think could also have some conflicts if it merges - so I may start thinking about what I have outstanding that can rebase *on top of* that one | 17:48 |
| acoles | I like that one | 17:49 |
| clayg | acoles: oh there's a bunch of goodn's out there | 17:49 |
| acoles | clayg: the only one I didn't have +2 on was patch 418691, IIRC there was an issue over handling None from read_hashes | 17:51 |
| patchbot | https://review.openstack.org/#/c/418691/ - swift - Fix performance regression with hash invalidations | 17:51 |
| clayg | yeah patch 418691 is the one that will get messed up by patch 419787 for *sure* | 17:53 |
| patchbot | https://review.openstack.org/#/c/418691/ - swift - Fix performance regression with hash invalidations | 17:53 |
| patchbot | https://review.openstack.org/#/c/419787/ - swift - Better optimistic lock in get_hashes | 17:53 |
| clayg | it's also the one that needs patch 422076 the most - and what I was talking about with letting the LockTimeout get out of consolidate_hashes | 17:54 |
| patchbot | https://review.openstack.org/#/c/422076/ - swift - Handle unreadable hashes.pkl same way as non-exist... | 17:54 |
| clayg | mainly I just want conslidate_hashes to either return with the "safe_inode" from hashes.pkl or raise an exeption that terminates get_hashes | 17:54 |
| clayg | I don't want to have to handle "any" exception by calling safe_get_inode outside of lockdir because consolidate_hashes blew up | 17:55 |
| *** david-lyle_ has joined #openstack-swift | 17:56 | |
| *** david-lyle_ has quit IRC | 17:56 | |
| *** Jeffrey4l__ has joined #openstack-swift | 17:57 | |
| *** david-lyle_ has joined #openstack-swift | 17:57 | |
| *** Jeffrey4l_ has quit IRC | 17:59 | |
| *** rledisez has quit IRC | 18:02 | |
| acoles | clayg: yep think I was starting to think that way - have consolidate_hashes always result in a hashes.pkl existing and if it fails then bail out | 18:02 |
| acoles | rather than dealing with the new partition case outside of it | 18:02 |
| clayg | acoles: I wasn't sure if I'd gotten as far as it should always create the hashes.pkl - but I like the implications | 18:03 |
| acoles | oic, yeah of course safe_get_inode can return None. well, IDK, I just started warning to the idea that consolidate_hashes would "just get me a consolidated hashes+hashes.pkl to work with" | 18:04 |
| acoles | warming* | 18:04 |
| acoles | but w/e, we're making progress! | 18:05 |
| *** david-lyle_ is now known as david-lyle | 18:05 | |
| clayg | yeah - i'm remembering I had some weird edge case where the hashes.pkl was invalid (EOFError/zero-byte/etc) but there wasn't any suffixes that would trigger the rewrite | 18:07 |
| clayg | and it would just stay there... invalid... | 18:07 |
| acoles | yes, if it exists but is unreadable for some reason | 18:08 |
| clayg | this was all in my mind tho - so I thought it may have just been a dream | 18:08 |
| *** dmorita has joined #openstack-swift | 18:08 | |
| clayg | yeah, basically the reason force_rewrite came to exist in the first place | 18:08 |
| acoles | readable but invalid returns None so we do a listdir etc, but exists and unreadable returns {}. | 18:08 |
| clayg | I don't want conslidate_hashes to have to return some indication that it wasn't able to read | 18:08 |
| acoles | wait, that was what patch 422076 was about | 18:09 |
| patchbot | https://review.openstack.org/#/c/422076/ - swift - Handle unreadable hashes.pkl same way as non-exist... | 18:09 |
| acoles | maybe i am getting confused | 18:09 |
| clayg | acoles: it still does "a dict or None" | 18:09 |
| zaitcev | I'm in your Swift cloning your MIMEputter | 18:09 |
| zaitcev | er | 18:10 |
| clayg | zaitcev: NOOOOooo - i thought you were going to fix the whole footer protocol thing!? | 18:10 |
| zaitcev | clayg: well yea, it's called DoublePutter now | 18:10 |
| clayg | remember!? the HTTP bug? And the "golang will never let their httpserver be as broken as we made eventlet.wsgi" | 18:10 |
| clayg | zaitcev: wahhhhhhaaaaaAAAAA? nice! | 18:11 |
| zaitcev | yes, duh. You told me to do it | 18:11 |
| clayg | to be 100% clear - I told you *not* to do the dumb thing - and when you asked "why didn't we do the smart thing" - I was like "yeah... because we're dumb" | 18:12 |
| acoles | clayg: yep, its the unreadable->empty dict that I was changing there | 18:12 |
| clayg | acoles: ok, so yeah - then maybe you already have the situation where if the os.listdir doesn't find any suffixes to add to the empty dict - it won't be modified and need to rewrite so it will continue to be invalid on disk!? | 18:13 |
| *** jordanP has quit IRC | 18:13 | |
| clayg | acoles: not saying it's bad - i just like the idea you may have independently come to the same conclusion as I did - makes it seem more "correct" | 18:13 |
| acoles | clayg: something like that, so if nothing changes you'll keep reporting hashes={} IIRC | 18:14 |
| acoles | well, until the deterministic 10th cycle | 18:14 |
| clayg | acoles: then what happens? I think empty hashes will always do the listdir - the question is if the listdir finds any suffixes to mark None set's modified to True and tries to rewrite it? | 18:14 |
| *** klrmn has joined #openstack-swift | 18:15 | |
| clayg | oic, do_listdir always set's modified to True | 18:15 |
| clayg | i forgot about that - it always looked dedented to me | 18:15 |
| clayg | oh, empty dict doesn't force listdir - that can't be correct for invalid - it needs to listdir | 18:16 |
| acoles | clayg: so, I wrote this test (patch 422076 depends on this patch 421385) https://review.openstack.org/#/c/421385/2/test/unit/obj/test_diskfile.py@7382 | 18:18 |
| patchbot | https://review.openstack.org/#/c/422076/ - swift - Handle unreadable hashes.pkl same way as non-exist... | 18:18 |
| patchbot | https://review.openstack.org/#/c/421385/ - swift - Add more tests for diskfile consolidate_hashes | 18:18 |
| patchbot | patch 421385 - swift - Add more tests for diskfile consolidate_hashes | 18:18 |
| acoles | clayg: it is somewhat contrived | 18:19 |
| acoles | but left me thinking why treat unreadable different from invalid or non-existent? we have to assume we have no idea what current hash state is in all cases | 18:20 |
| acoles | clayg: sorry, gotta dash. look forward to reviewing some updates :) | 18:20 |
| clayg | acoles: k, thanks for the pointers! | 18:21 |
| *** acoles is now known as acoles_ | 18:21 | |
| *** ChubYann has joined #openstack-swift | 18:25 | |
| *** geaaru has quit IRC | 18:40 | |
| *** silor has joined #openstack-swift | 18:44 | |
| *** catintheroof has joined #openstack-swift | 18:45 | |
| *** catinthe_ has quit IRC | 18:47 | |
| *** mingyu has joined #openstack-swift | 18:49 | |
| *** mingyu has quit IRC | 18:53 | |
| notmyname | I'm slightly worried about the openstack-wide discussions about how md5 can't be used (or must be wrapped) | 19:03 |
| *** catinthe_ has joined #openstack-swift | 19:03 | |
| notmyname | ie http://lists.openstack.org/pipermail/openstack-dev/2017-January/110278.html | 19:04 |
| *** catintheroof has quit IRC | 19:06 | |
| clayg | notmyname: i'm just worried about the current social setup and prioritization | 19:08 |
| timburke | notmyname: it put me in a mind to start thinking about how we could replace at least *some* of our use with other algorithms. say, create a new storage policy and have the ring inform hash_path to use sha256 or something | 19:08 |
| notmyname | yeah. like many people, I'd prefer to have not-md5. but I'm worried about API compatibility and code churn | 19:09 |
| clayg | notmyname: the idea that "someone" creates an abstraction around our use of md5 and eplicitly annoates useforsecurity=False is fine | 19:09 |
| timburke | clayg: agreed that that's the much more frightening thing. but i can *do* something about the hashes we use | 19:09 |
| clayg | there's *some* hope that someday that abstraction could be leveraged to allow in some future-hand-wave where we get sha256 for get_hash_path and crc for ETags - but of course the migration and backwards compat is the hard part - not the global s/hashlib/oslo/ | 19:10 |
| notmyname | https://blake2.net | 19:10 |
| clayg | but you know "every journey begins with a a step" is something people who like to do stuff say | 19:10 |
| timburke | notmyname: py36 has it built in... | 19:10 |
| notmyname | timburke: well there's the reason to do py3 ;-) | 19:11 |
| clayg | notmyname: I think that may be more "fast" in terms of throughput - there's other algos that go for "wide-even-distribution-of-short-strings-fast" | 19:11 |
| clayg | unless you want a cryptographically secure ETag (!?) | 19:12 |
| clayg | doesn't matter | 19:12 |
| clayg | that ML isn't about getting to *use* other hashes - it's about annotating the hashes we *are* using as useforsecurity=False | 19:12 |
| clayg | timburke: oh... or your saying we take the marching order for the audit as an oppertunity to shelve other stuff we're working on in favor of addressing this critical hash function question that's slowing swift adoption? | 19:14 |
| timburke | clayg: i'm saying that the conversation put an idea into my head. didn't say i wrote (or will write) any code about it :-P | 19:15 |
| clayg | notmyname: my worry as it were is that "someone" will be either a) an openstack goal and we end up doing it before we really "need" to because OpenStack or b) someone tries to cross-project and ends up mad at Swift because... of me mostly I guess | 19:16 |
| timburke | clayg: go back to *actually* addressing some critical hash-related bugs in swift ;-) | 19:16 |
| clayg | timburke: yessir! | 19:16 |
| *** ganders has quit IRC | 19:27 | |
| dims | timburke : quick question, does /info work under py3 yet? | 19:38 |
| dims | i've had to disable swift in many of the jobs for py35 | 19:39 |
| notmyname | dims: he's standing at my desk right now. he'll answer soon | 19:42 |
| dims | hi notmyname thanks! | 19:42 |
| *** vint_bra has quit IRC | 19:45 | |
| timburke | dims: ...maybe? with your patch to actually get the proxy server to start, it was probably willing to send back *some* response, and since /info is served entirely from the proxy there's a decent chance it wasn't a 500. that said, there probably would be some... problems... with the response. i think i remember seeing things like `Content-Length: b'1146'` in the headers the last time i looked into it | 19:47 |
| *** vint_bra has joined #openstack-swift | 19:47 | |
| *** McMurlock1 has joined #openstack-swift | 19:47 | |
| dims | timburke : swift generates the content-length info? | 19:48 |
| timburke | and it goes in and out of a HeaderKeyDict a few times as it comes back through the pipeline, which may or may not be doing the right thing | 19:50 |
| notmyname | dims: yes, we do (since headers are sent first before the body, and stuff returned from swift can be very large, we can't simply pass the response body back up to the wsgi server to calculate) | 19:58 |
| dims | gotcha | 19:59 |
| timburke | notmyname: well, we *could* for /info, but (to my knowledge) we don't, because we're so in the habit of supplying it ourselves | 20:00 |
| notmyname | dims: well, that's the general case. seems like we don't for...yeah. what timburke said | 20:00 |
| notmyname | well, we aren't explicitly providing it in the InfoController for /info responses, but we're passing it to swob (our webob replacement) and that's doing it | 20:01 |
| notmyname | hmm...or not. if we don't set it (we don't on /info responses) looks like we might just punt up to the wsgi server | 20:02 |
| notmyname | however, all that being said, timburke's comments about likely having to clean stuff up still stands | 20:04 |
| notmyname | #different_topic | 20:05 |
| notmyname | so we gotta do a swiftclient release this week | 20:06 |
| dims | is the /info the easiest path to try to get to work? notmyname timburke | 20:06 |
| timburke | notmyname: too bad! the info controller does it by way of swob.Response (because it set a body) https://github.com/openstack/swift/blob/master/swift/common/swob.py#L315-L321 | 20:06 |
| openstackgerrit | Thiago da Silva proposed openstack/swift: remove reference to deprecated tool https://review.openstack.org/424312 | 20:06 |
| timburke | dims: yeah, that's where i'd start | 20:06 |
| notmyname | dims: it has the fewest requirements. so yes | 20:06 |
| notmyname | timburke: right. that's what I was getting down to in my code search. | 20:08 |
| timburke | the next step would probably be something like making account requests work. the big pain is likely to be making https://github.com/openstack/swift/blob/master/swift/common/bufferedhttp.py work | 20:08 |
| notmyname | that's the internal socket stuff, right? | 20:08 |
| timburke | yeah, i still don't know how we're going to (properly) handle https://github.com/openstack/swift/blob/master/swift/common/bufferedhttp.py#L57 in py3 -- socket._socketobject flat out doesn't exist and i haven't had the time or inclination to try to find its replacement yet | 20:10 |
| timburke | on the swiftclient release, https://review.openstack.org/#/c/413348/ might be good, particularly ahead of the stable branching | 20:17 |
| patchbot | patch 413348 - python-swiftclient - Add Constraints support | 20:17 |
| timburke | https://review.openstack.org/#/c/359477/ is a nice developer-facing usability improvement | 20:18 |
| patchbot | patch 359477 - python-swiftclient - Accept more types of input for headers/meta | 20:18 |
| dims | and i have no clue about the code base :) | 20:18 |
| timburke | dims: you and me both. i at least have an idea of what it's trying to do, but i'm not sure how to validate that it still works, nor am i familiar enough with the stdlib to figure out what the modern equivalent is | 20:20 |
| *** silor has quit IRC | 20:24 | |
| timburke | https://review.openstack.org/#/c/310075/ would be nice, but i have to think more about how to do it -- currently, i think every POST will require a HEAD :-/ | 20:25 |
| patchbot | patch 310075 - python-swiftclient - Expire segments when expiring large objects | 20:25 |
| dims | notmyname : timburke : what's the simplest curl command that you use? | 20:25 |
| timburke | curl -v http://swift-host:8080/info | 20:26 |
| timburke | there's no auth required | 20:26 |
| dims | ack thanks | 20:27 |
| timburke | https://bugs.launchpad.net/python-swiftclient/+bug/1586690 and https://bugs.launchpad.net/python-swiftclient/+bug/1621581 are strange -- i'm not sure they're client bugs? seems like some problem with the server config | 20:27 |
| openstack | Launchpad bug 1586690 in python-swiftclient "Uploading empty(0 B) file fails" [Undecided,Incomplete] - Assigned to Uday Swami (swamius) | 20:27 |
| openstack | Launchpad bug 1621581 in python-swiftclient "swiftclient returns response headers without 'Content-Length' param, thus causing upload object to fail" [Undecided,In progress] - Assigned to Arun Mani (arun-mani) | 20:27 |
| timburke | https://review.openstack.org/#/c/297958/ is kinda interesting; i never got around to following up again... | 20:29 |
| patchbot | patch 297958 - python-swiftclient - Check /info before send SLO segments | 20:29 |
| timburke | https://review.openstack.org/#/c/259410/ might be nice, but i think we need to think harder about how much we buffer and where we buffer it | 20:32 |
| patchbot | patch 259410 - python-swiftclient - Support uploading to an object in swift from stdin | 20:32 |
| timburke | and i still need to give a response to joel on https://review.openstack.org/#/c/399756/ | 20:32 |
| patchbot | patch 399756 - python-swiftclient - On auth failure, ignore passed-in tokens and retry | 20:32 |
| timburke | may as well land https://review.openstack.org/#/c/360135/ | 20:32 |
| patchbot | patch 360135 - python-swiftclient - Make functests py3-compatible | 20:32 |
| clayg | timburke: we should just update bufferedhttp to keep a reference to "real_close" https://github.com/python/cpython/blob/master/Lib/socket.py#L409 | 20:48 |
| clayg | getting ahold of something that will "screw your ref count and close this ^&*" is the only reason we ever wanted _real_socket in the first place | 20:49 |
| *** mingyu has joined #openstack-swift | 20:50 | |
| timburke | clayg: either way, someone still needs to invest the time in finding the py3 equivalent and validating that it really does do the thing we need it to | 20:52 |
| clayg | not 100% clear to me if knowing what to do in that case acctually helps anything - it's not like I'm acctually going to take up the py3 cause anytime soon that I'm aware of | 20:52 |
| clayg | timburke: reading the close method just below it's pretty clear that *it* thinks _real_close does the "really acctually close the socket now" dance - which I can confirm is *why* we grew the reference to _real_sock and what we want it to do | 20:53 |
| clayg | someone *could* change it "PY3: blah blah" and as long as py2 still calls _real_socket.close() i'm satisified we don't need to go out of our want to track down the reproduce case | 20:54 |
| *** mingyu has quit IRC | 20:54 | |
| timburke | bah, i'd glossed over the python source link -- yeah, *probably* does what it says on the tin? | 20:55 |
| clayg | but +1 to not signing up to do it! But if someone *asks* - i'd point them that way. And if they asked to review a patch that did that - I'd probably sanity the regression on py2 and be happy to merge | 20:56 |
| clayg | timburke: i mean... the next stop would be socket.c in modules or something i guess - see what close acctually does - but heavn help of if it's not a thin shim around man -s 2 close | 20:56 |
| mattoliverau | Morning | 20:57 |
| clayg | anyway... if that *doesn't* work in py3 it's NBD - I mean so py3 grows a socket leak regression ... but only once someone gets py3 working and deployed in production (where'd you care) | 20:57 |
| clayg | it doesn't risk eating your data - so at that point if someone files the bug "leaked socket in py3" we can kick ourselves for not doing it better | 20:58 |
| clayg | doesn't like... swift.conf not work in py3 or something too? | 20:58 |
| clayg | mattoliverau: should I watch the container sharding talk from LCA!? | 20:58 |
| mattoliverau | clayg: if you need help falling to sleep then sure :p it's just a high level overview of the process of sharding, but could help if you can't remember some of the discussions | 21:02 |
| timburke | mattoliverau: linky? | 21:04 |
| mattoliverau | I plan to write up some decent documentation by the time the PTG at the latest. Whick will have details of all the design, include interesting internals. | 21:05 |
| mattoliverau | Which will help y'all understand and then make my code better ;) | 21:06 |
| clayg | I was just going to google "how to put a really big thing in your cluster by the man down under" | 21:06 |
| mattoliverau | Lol | 21:08 |
| mattoliverau | youtu.be/g7tSqRNFaO0 | 21:08 |
| clayg | you sound like *extra* austrailian when you're in austraila I think | 21:10 |
| clayg | ;) | 21:10 |
| mattoliverau | Oh really, I always thought I was extra Australian when outside oz | 21:10 |
| mattoliverau | I didn't like the intro.. "this Rackspace thing".. when all she needed do was read the title :( | 21:11 |
| clayg | mattoliverau: i'm probably jsut *missing* you | 21:11 |
| *** jerrygb has quit IRC | 21:12 | |
| timburke | oooh and i see some new, pretty graphs on https://oliver.net.au/?p=285 | 21:13 |
| clayg | mattoliverau: it's all good - only way it woulda been better is if she'd have called you Swift PTL! | 21:13 |
| *** JimCheung has quit IRC | 21:13 | |
| *** JimCheung has joined #openstack-swift | 21:14 | |
| mattoliverau | Lol yeah. Oh well, at least I got to speak about it :) | 21:15 |
| mattoliverau | timburke: yeah, need to change my blogs theme cause graphs images don't display quite right. | 21:16 |
| mattoliverau | If you want to see more LCA videos: https://www.youtube.com/channel/UCDMo9DyACXG62ak5cVgs3TA | 21:19 |
| *** jerrygb has joined #openstack-swift | 21:31 | |
| *** jerrygb_ has joined #openstack-swift | 21:32 | |
| openstackgerrit | Davanum Srinivas (dims) proposed openstack/swift: [WIP] Get swift process to come up under py35 https://review.openstack.org/416084 | 21:33 |
| *** Jeffrey4l_ has joined #openstack-swift | 21:34 | |
| *** Jeffrey4l__ has quit IRC | 21:35 | |
| *** jerrygb has quit IRC | 21:36 | |
| *** mingyu has joined #openstack-swift | 21:51 | |
| dims | timburke : so...got /info working http://paste.openstack.org/show/596153/ | 21:53 |
| timburke | dims: sweet! | 21:55 |
| *** mingyu has quit IRC | 21:56 | |
| notmyname | I'm guessing there will be an openstack summit in Chicago https://twitter.com/OpenStack/status/823630404468908032 | 21:56 |
| notmyname | oops. boston | 21:56 |
| dims | guess this is as far i can get with the _socketobject stuff you mentioned... | 21:56 |
| * notmyname might not keep up with baseball | 21:56 | |
| notmyname | so maybe a thing at the boston summit | 21:56 |
| timburke | dims: clayg seemed to have some ideas involving socket._real_close ? if you're interested | 21:57 |
| dims | timburke : i will check back in a few days :) hoping y'all get a little more further by then | 21:58 |
| clayg | timburke: don't put that on me! I don't even know what we're trying to solve for! | 21:58 |
| dims | updated https://review.openstack.org/#/c/416084/ | 21:58 |
| patchbot | patch 416084 - swift - [WIP] Get swift process to come up under py35 | 21:58 |
| notmyname | mattoliverau: it might be nice to share the "ugly" graphs you have. the ones where the VM noisy neighbor hosed you. wasnt' there some good info before that happened? | 21:58 |
| dims | LOL clayg | 21:58 |
| dims | clayg : simple, run everything under python3.5 :) | 21:58 |
| timburke | run *correctly*. important distinction | 21:59 |
| clayg | right ... | 21:59 |
| clayg | timburke: OMG! you may have cracked it? Is "correctly" part of the goal? | 21:59 |
| dims | :) | 21:59 |
| timburke | if six.PY3: return HTTPNotImplemented() | 22:00 |
| dims | timburke : so we have a CI job as of this morning in nova check queue that runs neutron/glance/cinder/nova everything under python3.5 and runs the tempest test suite :) | 22:00 |
| dims | so yes, +1 to "correctly" :) | 22:00 |
| *** vint_bra has quit IRC | 22:01 | |
| timburke | dims: i'm still waiting on https://review.openstack.org/#/c/401397/ so... we'll see? i'm afraid we haven't had a great track record WRT to py3 review turn-around time | 22:01 |
| patchbot | patch 401397 - swift - py3: port common/ring/, common/linkat.py, and comm... | 22:01 |
| dims | ack timburke | 22:01 |
| timburke | hopefully we'll be a bit further | 22:01 |
| *** McMurlock1 has quit IRC | 22:01 | |
| mattoliverau | notmyname: sure, I can do that thing :) | 22:02 |
| clayg | "slicing" | 22:02 |
| mattoliverau | lol, yeah, this is why notmyname should come up with names.. and not me ;) | 22:02 |
| timburke | dims: i think the ability to actually *do something* with py3 is a step in the right direction, though. just porting unit tests isn't likely to motivate, in no small art because we have a hard time verifying that we're still actually testing the right things | 22:03 |
| dims | right, so i first got the nova functional tests working, then rally test and then graduated to tempest | 22:04 |
| *** sams-gleb has quit IRC | 22:05 | |
| timburke | fwiw, it may be easier to start with a py3 proxy, py2 everything else | 22:06 |
| notmyname | timburke: dims: ...especially when considering project plans to delete the python object server. so not a lot of motivation to port that to py3 | 22:06 |
| notmyname | so I'd echo the "start with proxy server" part | 22:07 |
| timburke | more than that; py3 proxy talking to py3 account server would be nice to have later, but a lot to try to debug right off the bat | 22:08 |
| openstackgerrit | Jim Cheung proposed openstack/liberasurecode: Add Phazr.IO libphazr backend to liberasurecode https://review.openstack.org/424353 | 22:13 |
| clayg | "bahrrhaahahhh crazy" | 22:19 |
| notmyname | JimCheung: cool. nice to see the patch up in gerrit :-) | 22:20 |
| JimCheung | Thanks notmyname: I did make a booboo and didn't include a brief description of the changes tho | 22:20 |
| JimCheung | Any easy way to mend it? | 22:21 |
| notmyname | JimCheung: well, yeah, I noticed that too :-) | 22:21 |
| JimCheung | :-) | 22:21 |
| *** mmotiani has joined #openstack-swift | 22:22 | |
| notmyname | JimCheung: there *used* to be an "edit commit message" button, but I don't see that now. you can also amend the commit locally and then `git review` again | 22:22 |
| notmyname | JimCheung: just keep the Change-Id the same, and you'll be good | 22:22 |
| JimCheung | Okay, thanks! | 22:23 |
| openstackgerrit | Jim Cheung proposed openstack/liberasurecode: Add Phazr.IO libphazr backend to liberasurecode https://review.openstack.org/424353 | 22:26 |
| openstackgerrit | Jim Cheung proposed openstack/liberasurecode: Add Phazr.IO libphazr backend to liberasurecode https://review.openstack.org/424353 | 22:28 |
| JimCheung | Comments added. Thanks! | 22:29 |
| notmyname | thanks | 22:29 |
| clayg | what have we done :'( | 22:32 |
| notmyname | clayg: with what where? | 22:33 |
| timburke | clayg: accidentally become C library maintainers? | 22:33 |
| notmyname | timburke: lol | 22:33 |
| clayg | worse! An opensource C library that's just a plugin wrapper to other interesting libraries and some of them may not even be FOSS! | 22:34 |
| notmyname | true, phazr isn't FOSS. but the fact that we've got people who want to add support for more EC types is really great | 22:35 |
| timburke | JimCheung: is there any way for us to run tests with libphazr enabled? we've already got an example of a backend where we can't do that, but it'll make the change easier to review i think | 22:36 |
| timburke | oh, or i need to read the updated commit message, then contact support@ :-) | 22:37 |
| JimCheung | We can submit a binary to help testing if needed | 22:38 |
| JimCheung | :-) | 22:38 |
| notmyname | JimCheung: yeah, it's going to be important to be able to test the code in order to accept it and maintain it | 22:38 |
| JimCheung | We'll figure out something quick | 22:39 |
| JimCheung | How did you test SHSS? | 22:40 |
| notmyname | yeah, I was about to type soemthing about that :-) | 22:41 |
| JimCheung | The submitted changes are almost the same as SHSS. | 22:41 |
| notmyname | honestly I wonder if it's just because kota_ is active with libec | 22:41 |
| notmyname | however, that's not a long-term good thing | 22:41 |
| notmyname | I'd definitely like to get his input, but it's still pretty early in tokyo | 22:42 |
| JimCheung | Sounds good | 22:42 |
| mattoliverau | also kota_ was sick yesterday (according to the twitter) | 22:43 |
| notmyname | yeah, I saw that too | 22:43 |
| *** mingyu has joined #openstack-swift | 22:52 | |
| *** mingyu has quit IRC | 22:56 | |
| *** sams-gleb has joined #openstack-swift | 23:06 | |
| *** caiobrentano_ has quit IRC | 23:07 | |
| *** remix_tj has quit IRC | 23:07 | |
| *** remix_tj has joined #openstack-swift | 23:09 | |
| *** sams-gleb has quit IRC | 23:12 | |
| kota_ | Still a high fever, sorry | 23:16 |
| notmyname | kota_: hope you feel better soon | 23:17 |
| notmyname | kota_: please don't feel any pressure to do upstream stuff. spend time getting well | 23:17 |
| kota_ | Thx, will take a look the log | 23:17 |
| kota_ | Later | 23:18 |
| *** catinthe_ has quit IRC | 23:23 | |
| *** chsc has quit IRC | 23:38 | |
| *** kei_yama has joined #openstack-swift | 23:46 | |
| *** caiobrentano_ has joined #openstack-swift | 23:52 | |
| openstackgerrit | OpenStack Proposal Bot proposed openstack/swift: Updated from global requirements https://review.openstack.org/88736 | 23:53 |
Generated by irclog2html.py 2.14.0 by Marius Gedminas - find it at mg.pov.lt!