| timburke | oh extra fun, even openstackgerrit is gone... | 00:00 |
|---|---|---|
| timburke | i knew i shoulda gone and downloaded the undelete patch as i was reviewing... | 00:02 |
| zaitcev | yeah, always git review -d N | 00:24 |
| *** linkmark has quit IRC | 00:25 | |
| zaitcev | oh, crap | 00:25 |
| zaitcev | Programming in Go is harmful, you start returning tuples everywhere. | 00:25 |
| notmyname | zaitcev: do you know if you'll be able to go to dublin for the PTG? | 00:25 |
| zaitcev | notmyname: Most likely yes, although I haven't gotten air tickets yet. | 00:26 |
| notmyname | great! | 00:26 |
| zaitcev | notmyname: I addressed all issues raised at the big PUT+POST video meeting, so I'm going to poke Clay et al for reviews before PTG. Then maybe PTG will only need to rubber-stamp it, if that. We committed Symlink without in-person meetings, so surely it can be done. | 00:27 |
| zaitcev | Man, I really wish Mr. Gun1x saved that account.builder that makes a traceback just from looking at it. | 00:29 |
| zaitcev | Looked like EmptyRingError wasn't caught as intended | 00:29 |
| zaitcev | okay, I need to continue setting up a TV that I bought just now. | 00:29 |
| *** itlinux has joined #openstack-swift | 00:43 | |
| *** tovin07_ has joined #openstack-swift | 00:58 | |
| *** cshastri_ has joined #openstack-swift | 01:33 | |
| *** awkwardpenguin has joined #openstack-swift | 01:42 | |
| *** psachin has joined #openstack-swift | 01:47 | |
| *** awkwardpenguin has quit IRC | 02:03 | |
| *** openstackgerrit has joined #openstack-swift | 02:27 | |
| openstackgerrit | Merged openstack/swift master: Better mock out OSErrors in test_replicator before raising them https://review.openstack.org/496433 | 02:27 |
| *** bkopilov has quit IRC | 03:14 | |
| *** kei_yama has quit IRC | 03:45 | |
| *** awkwardpenguin has joined #openstack-swift | 03:46 | |
| *** awkwardpenguin has quit IRC | 03:46 | |
| *** awkwardpenguin has joined #openstack-swift | 03:48 | |
| *** kei_yama has joined #openstack-swift | 04:02 | |
| *** links has joined #openstack-swift | 04:12 | |
| *** kei_yama has quit IRC | 04:12 | |
| openstackgerrit | Tim Burke proposed openstack/swift master: tempurl: Make the digest algorithm configurable https://review.openstack.org/525770 | 04:20 |
| openstackgerrit | Tim Burke proposed openstack/swift master: tempurl: Deprecate sha1 signatures https://review.openstack.org/525771 | 04:20 |
| *** bkopilov has joined #openstack-swift | 04:33 | |
| *** kei_yama has joined #openstack-swift | 04:48 | |
| *** gkadam has joined #openstack-swift | 05:23 | |
| *** gkadam has quit IRC | 05:24 | |
| *** gkadam has joined #openstack-swift | 05:24 | |
| *** gkadam has quit IRC | 06:01 | |
| *** gkadam has joined #openstack-swift | 06:01 | |
| *** kei_yama has quit IRC | 06:04 | |
| *** itlinux has quit IRC | 06:08 | |
| *** gkadam has quit IRC | 06:25 | |
| *** gkadam has joined #openstack-swift | 06:26 | |
| *** SkyRocknRoll has joined #openstack-swift | 06:45 | |
| *** awkwardpenguin has quit IRC | 07:03 | |
| *** awkwardpenguin has joined #openstack-swift | 07:04 | |
| *** gkadam has quit IRC | 07:06 | |
| *** awkwardpenguin has quit IRC | 07:09 | |
| *** gkadam has joined #openstack-swift | 07:25 | |
| *** gkadam has quit IRC | 07:50 | |
| *** awkwardpenguin has joined #openstack-swift | 09:06 | |
| *** awkwardpenguin has quit IRC | 09:10 | |
| openstackgerrit | Guoqiang Ding proposed openstack/python-swiftclient master: Update docs build job for new documentation PTI jobs https://review.openstack.org/530330 | 09:37 |
| *** silor has joined #openstack-swift | 09:40 | |
| *** tovin07_ has quit IRC | 10:12 | |
| *** psachin has quit IRC | 10:21 | |
| *** cshastri_ has quit IRC | 10:25 | |
| gun1x | notmyname: i managed to get this done | 10:28 |
| gun1x | i just reinstalled the storage node | 10:28 |
| gun1x | zaitcev: yea, sorry, i wanted to get this done fast :D | 10:29 |
| *** awkwardpenguin has joined #openstack-swift | 11:08 | |
| *** awkwardpenguin has quit IRC | 11:12 | |
| *** silor has quit IRC | 11:36 | |
| *** bkopilov has quit IRC | 12:09 | |
| *** hoonetorg has quit IRC | 12:27 | |
| *** hoonetorg has joined #openstack-swift | 12:56 | |
| *** hoonetorg_ has joined #openstack-swift | 12:57 | |
| *** hoonetorg_ has quit IRC | 12:58 | |
| *** cbartz has joined #openstack-swift | 13:00 | |
| *** awkwardpenguin has joined #openstack-swift | 13:09 | |
| *** awkwardpenguin has quit IRC | 13:14 | |
| *** hoonetorg has quit IRC | 13:17 | |
| *** hoonetorg has joined #openstack-swift | 13:33 | |
| *** gkadam has joined #openstack-swift | 13:37 | |
| *** links has quit IRC | 13:57 | |
| *** silor has joined #openstack-swift | 14:00 | |
| *** bkopilov has joined #openstack-swift | 14:15 | |
| *** awkwardpenguin has joined #openstack-swift | 14:33 | |
| *** awkwardpenguin has quit IRC | 14:37 | |
| *** gkadam has quit IRC | 14:41 | |
| *** itlinux has joined #openstack-swift | 15:05 | |
| *** silor has quit IRC | 15:09 | |
| *** itlinux has quit IRC | 15:12 | |
| openstackgerrit | Guoqiang Ding proposed openstack/python-swiftclient master: Update new documentation PTI jobs https://review.openstack.org/530330 | 15:25 |
| *** SkyRocknRoll has quit IRC | 15:27 | |
| *** SkyRocknRoll has joined #openstack-swift | 15:36 | |
| *** SkyRocknRoll has quit IRC | 15:41 | |
| *** SkyRocknRoll has joined #openstack-swift | 15:53 | |
| *** itlinux has joined #openstack-swift | 16:03 | |
| openstackgerrit | Guoqiang Ding proposed openstack/python-swiftclient master: Update new documentation PTI jobs https://review.openstack.org/530330 | 16:16 |
| *** awkwardpenguin has joined #openstack-swift | 16:35 | |
| *** SkyRocknRoll has quit IRC | 16:35 | |
| *** awkwardpenguin has quit IRC | 16:39 | |
| *** cbartz has quit IRC | 16:49 | |
| *** SkyRocknRoll has joined #openstack-swift | 16:52 | |
| timburke | good morning | 16:52 |
| notmyname | good morning | 17:29 |
| *** two_tired has joined #openstack-swift | 17:42 | |
| clayg | timburke: thanks for all the great comments on patch 528155 - I'll see what I can do! | 17:55 |
| patchbot | https://review.openstack.org/#/c/528155/ - swift - Represent dispersion worse than one replicanth | 17:55 |
| timburke | yeah! i was just about to ping you on that. i'm pretty sure deducting max_replicas is the right move there -- i was suspicious about how big some of those max_dispersion numbers were | 17:57 |
| timburke | i've got a diff stashed somewhere... gimme a sec... | 17:57 |
| timburke | clayg: see what you think of http://paste.openstack.org/show/632654/ | 17:58 |
| clayg | so if a tier should only have at max 1 replica of a part per maximum dispersion - and it has *two* - then only what part-replica is at risk... | 18:01 |
| clayg | i guess - does that make everyone's dispersion like a ton smaller? I was already weirded out at how smaller our dispersion numbers are getting... and I think it's weird that the more replicas you have the more you have to worry about even a tiny bit of dispersion being really rather bad | 18:02 |
| timburke | maybe i'd be ok taking out the divided-by-total-replicas so overall dispersion can go past 100? but then i'd have to think again about what's going on in the per-tier calculations... | 18:05 |
| clayg | yeah i have no idea :'( | 18:05 |
| clayg | it's all made up | 18:05 |
| clayg | and then we're going to make it up differently now | 18:05 |
| clayg | but ... there's got to be some benifit in having 3-4 people be like "yeah, that number sort of makes some intuitive sense to me" | 18:06 |
| clayg | notmyname just says it has to get bigger when it's badder | 18:06 |
| clayg | ^ which seems reasonable to me | 18:06 |
| notmyname | oh did I? | 18:07 |
| clayg | yeah because of balance working that way? | 18:07 |
| notmyname | I thought bigger numbers were always better ;-) | 18:07 |
| clayg | if you have two gauges - they should both go red in the same direction | 18:07 |
| notmyname | yeah, I definitely agree with that analogy | 18:07 |
| clayg | yeah I think you would have preferred if balance 100% meant 100% good-to-go but that wasn't up to us | 18:07 |
| clayg | so... 0-0 means good | 18:08 |
| notmyname | "dispersion" seems like a good thing (gut reaction). so bigger dispersion == better | 18:08 |
| clayg | sigh - you can't change your mind now (although mattoliverau made the same observation, honestly I'm pretty sure my first pass worked that way) | 18:08 |
| notmyname | lol | 18:08 |
| notmyname | TBH, I don't htink it matters too much. going in the same direction (and showing changes after the user does something) is way more important than which way it goes | 18:09 |
| clayg | I'm totally open to changing it! What I really want to be able to fix is rebalance of rings that so fucked dispersion is getting stuck at 100% terrible - I need a better measuring stick is all | 18:09 |
| clayg | I can start shitting out a boolean - "you_are_less_fucked_keep_going = True" | 18:09 |
| notmyname | I can be totally fine with the best number being 0.0. or the best number being 100. either way, I *really* like the goal behind the patch--give users a better measuring stick so they know to keep going | 18:11 |
| notmyname | $ swift-ring-builder object.builder rebalance | 18:11 |
| notmyname | Good job! Keep going! | 18:12 |
| notmyname | hmm... perhaps that's a little too patronizing ;-) | 18:12 |
| *** awkwardpenguin has joined #openstack-swift | 18:12 | |
| *** gkadam has joined #openstack-swift | 18:15 | |
| *** chsc has joined #openstack-swift | 18:38 | |
| *** chsc has joined #openstack-swift | 18:38 | |
| *** gkadam has quit IRC | 18:54 | |
| openstackgerrit | Merged openstack/swift master: Skip symlink + vw functional tests if symlink is not enabled https://review.openstack.org/530014 | 18:55 |
| *** cbartz has joined #openstack-swift | 19:00 | |
| *** cbartz has quit IRC | 19:01 | |
| timburke | i wonder if the dispersion report should grow some knowledge about the underlying storage policy... i mean, you can square dispersion *and* balance for a 10+3 EC policy with 3 zones that have 2 equally-weighted nodes and a fourth zone with a single half-weight node (assuming you have enough drives) -- but that doesn't mean you're in a good place! | 19:08 |
| timburke | i feel like somewhere we should be saying "good job standing up that extra zone -- but you've *still* likely got an availability issue if you lose two nodes! | 19:09 |
| openstackgerrit | Tim Burke proposed openstack/swift master: tempurl: Make the digest algorithm configurable https://review.openstack.org/525770 | 19:12 |
| openstackgerrit | Tim Burke proposed openstack/swift master: tempurl: Deprecate sha1 signatures https://review.openstack.org/525771 | 19:12 |
| *** awkwardpenguin has joined #openstack-swift | 19:14 | |
| timburke | notmyname: are the tempurl tests part of defcore? i'm wondering how much trouble it'll be to get that second one to pass (needs tempest updates as written) | 19:15 |
| notmyname | I think so. but let me check | 19:15 |
| notmyname | tempest.api.object_storage.test_object_temp_url.ObjectTempUrlTest.test_get_object_using_temp_url | 19:17 |
| openstackgerrit | Clay Gerrard proposed openstack/swift master: Represent dispersion worse than one replicanth https://review.openstack.org/528155 | 19:17 |
| notmyname | tempest.api.object_storage.test_object_temp_url.ObjectTempUrlTest.test_put_object_using_temp_url | 19:17 |
| notmyname | those two tests are in the defcore guidelines | 19:17 |
| notmyname | found on https://refstack.openstack.org/#/guidelines | 19:17 |
| notmyname | select the object storage target program | 19:17 |
| notmyname | not that tempest puts a uuid on each test (I think each time it's updated) and the defcore guidelines refer to a particular uuid | 19:18 |
| *** awkwardpenguin has quit IRC | 19:18 | |
| notmyname | you can see them under the "test list" link on the right | 19:18 |
| timburke | yeah, i'd found the tests... wasn't clear from there whether it was in or not. so... it'll probably be a while before we can drop sha1 from default allowed-signatures | 19:28 |
| notmyname | nothing different than our normal deprication policy, I think | 19:32 |
| *** flwang has quit IRC | 20:48 | |
| *** flwang has joined #openstack-swift | 21:01 | |
| openstackgerrit | Merged openstack/swift master: add name to core emeritus https://review.openstack.org/527845 | 21:09 |
| clayg | timburke: there is not _currently_ anyway to recalculate dispersion w/o a rebalance | 21:20 |
| clayg | timburke: ultimately it happens in builder.RingBuilder._build_dispersion_graph which only gets called at the end of rebalance (and also dynamically if your on-disk serialized builder is missing the cached _dispersion_graph attribute because your it was balanced before we added that data) | 21:21 |
| clayg | timburke: I think it'd be a nice followup to add something like that... dispersion --recalculate (?) and that will also save back? dunno... | 21:22 |
| clayg | i think if we could agree on the interface the code would be trivial | 21:22 |
| clayg | thanks for the final review +2 on patch 528155 | 21:23 |
| patchbot | https://review.openstack.org/#/c/528155/ - swift - Represent dispersion worse than one replicanth | 21:23 |
| *** Jeffrey4l has quit IRC | 21:36 | |
| *** jlvillal has quit IRC | 21:37 | |
| *** jlvillal has joined #openstack-swift | 21:38 | |
| *** timur has quit IRC | 21:39 | |
| *** timur has joined #openstack-swift | 21:40 | |
| *** lifeless has quit IRC | 21:43 | |
| *** Jeffrey4l has joined #openstack-swift | 21:45 | |
| *** lifeless has joined #openstack-swift | 22:22 | |
| *** itlinux has quit IRC | 22:28 | |
| mattoliverau | I'm out at a holiday house with in-laws.. so looking forward to catching up with the dispersion patch whether it's landed or not when I'm back next week. (If I can't get to it before) | 22:51 |
| notmyname | mattoliverau: nice! enjoy the vacation. must be so good to get away from the beach house from time to time ;-) | 22:52 |
| notmyname | never have to make those hard decisions like "should I go walk on the beach now? or later? or both?" | 22:52 |
| *** itlinux has joined #openstack-swift | 23:16 | |
| *** awkwardpenguin has joined #openstack-swift | 23:22 | |
| openstackgerrit | Samuel Merritt proposed openstack/swift master: Fix socket leak on 416 EC GET responses. https://review.openstack.org/530411 | 23:26 |
| *** awkwardpenguin has quit IRC | 23:29 | |
| *** itlinux has quit IRC | 23:29 | |
| *** chsc has quit IRC | 23:32 | |
| clayg | # no super(); we inherit from an old-style class (it's | 23:37 |
| clayg | # httplib's fault; don't try and fix it). | 23:37 |
| clayg | ^ lol | 23:37 |
| *** SkyRocknRoll has quit IRC | 23:56 | |
Generated by irclog2html.py 2.15.3 by Marius Gedminas - find it at mg.pov.lt!