| *** mvalsecc has joined #openstack-swift | 00:09 | |
| *** gyee has quit IRC | 01:38 | |
| *** rcernin has quit IRC | 02:22 | |
| *** rcernin has joined #openstack-swift | 02:26 | |
| *** ianychoi has joined #openstack-swift | 04:36 | |
| *** m75abrams has joined #openstack-swift | 04:52 | |
| openstackgerrit | Pete Zaitcev proposed openstack/swift master: proxy: stop sending frags to PyECLib with a Queue https://review.opendev.org/704892 | 04:53 |
|---|---|---|
| *** evrardjp has quit IRC | 05:33 | |
| *** evrardjp has joined #openstack-swift | 05:33 | |
| *** m75abrams has quit IRC | 07:03 | |
| *** m75abrams has joined #openstack-swift | 07:04 | |
| *** rpittau|afk is now known as rpittau | 08:03 | |
| *** baojg has quit IRC | 08:34 | |
| *** baojg has joined #openstack-swift | 08:35 | |
| *** baojg has quit IRC | 08:35 | |
| *** rcernin has quit IRC | 08:37 | |
| *** baojg has joined #openstack-swift | 08:42 | |
| *** mvalsecc has quit IRC | 09:12 | |
| openstackgerrit | Alistair Coles proposed openstack/swift master: ssync: don't log tracebacks for client disconnects https://review.opendev.org/763205 | 10:14 |
| *** baojg has quit IRC | 11:04 | |
| *** baojg has joined #openstack-swift | 11:05 | |
| *** dsariel has joined #openstack-swift | 11:25 | |
| *** tkajinam has quit IRC | 11:29 | |
| *** tkajinam has joined #openstack-swift | 11:30 | |
| *** baojg has quit IRC | 11:40 | |
| *** baojg has joined #openstack-swift | 11:41 | |
| *** baojg has quit IRC | 11:55 | |
| *** baojg has joined #openstack-swift | 11:56 | |
| *** camelCaser has quit IRC | 12:15 | |
| *** rpittau is now known as rpittau|brb | 13:16 | |
| *** baojg has quit IRC | 14:36 | |
| *** baojg has joined #openstack-swift | 14:37 | |
| *** rpittau|brb is now known as rpittau | 14:51 | |
| *** m75abrams has quit IRC | 15:53 | |
| acoles | clayg: thanks for review of caching shard ranges, FYI I have squashed your test changes in | 16:15 |
| *** evrardjp has quit IRC | 16:17 | |
| openstackgerrit | Clay Gerrard proposed openstack/swift master: Enforce response length harder https://review.opendev.org/763405 | 16:20 |
| clayg | @acoles THANK YOU! my test "changes" were entirely mechanical - i genuinely found it helpful to break things up during my review when I was "zooming in" on a specific behaviors - I hope it didn't come across as entirely cosmetic nonsense busy work (it certainly WAS nit-picky!) | 16:25 |
| acoles | nah I had it in mind to refactor them so you just helped me along | 16:26 |
| acoles | sorry you had to grok the loooong version | 16:26 |
| clayg | yay i was helpful for a change!!! 🤪 | 16:30 |
| *** gyee has joined #openstack-swift | 16:34 | |
| *** gyee has quit IRC | 16:50 | |
| *** rpittau is now known as rpittau|afk | 16:52 | |
| *** gyee has joined #openstack-swift | 17:05 | |
| acoles | lol | 17:11 |
| *** klamath_atx has joined #openstack-swift | 17:20 | |
| klamath_atx | Good day, I was wondering when container sharder will be production ready. I noticed in the config examples it is listed as experimental however we have been using it with swiftstack for a while without issue. | 18:03 |
| *** dsariel has quit IRC | 18:05 | |
| *** dsariel has joined #openstack-swift | 18:06 | |
| timburke | klamath_atx, i think we'd probably feel comfortable removing that "experimental" tag once we've got a better handle on how to fix it when it goes off the rails (specifically, when there are overlapping shard ranges) | 18:06 |
| timburke | i'd say it's in a state like EC was for a while after it was released: it's certainly "production ready" in the sense that people are running it in production (you, me, a bunch of other former-swiftstack customers) | 18:08 |
| timburke | there's just a lot of rough edges out there that operators ought to be aware of if they're going to be using it | 18:09 |
| *** dsariel has quit IRC | 18:10 | |
| timburke | specifically: 1. fixing overlapping shard ranges (there's been a decent bit of work around this recently; it's kind of a big deal since sharding basically locks up for a container when it happens) | 18:11 |
| timburke | 2. we need a way to manually mark shards as shrinking, similar to what operators can do today to split up a large DB (there's some hope that this will fall out of the work for 1) | 18:12 |
| timburke | 3. we need a robust way to shard that won't lead to overlapping ranges (or, alternatively, we need to find a way to get the fix for 1 baked into the sharder process instead of requiring manual intervention) | 18:14 |
| timburke | there are a handful of other improvements that would be nice to have (things like caching shard ranges in more places; see p 761659), but those are the items i'd look for before removing the "experimental" tag | 18:16 |
| patchbot | https://review.opendev.org/#/c/761659/ - swift - WIP - Use cached shard ranges for container GETs - 4 patch sets | 18:16 |
| clayg | wait, is gerrit down or slow for anyone else? | 18:18 |
| clayg | timburke: I think if you're manually picking your replicas to run `swift-manage-shard-ranges` overlaps aren't as much of a problem - the reason *we* got overlaps was a bug in the controller's auto-sharding (which has since been fixed) | 18:19 |
| clayg | there's really nothing "experimental" about sharding - the problem is *auto* sharding - and the lack of shrinking | 18:20 |
| clayg | people suffering with big containers deserve to have - at least - the same confidence we have in `swift-manage-shard-ranges` to split up their worst offenders and move on with live | 18:20 |
| timburke | clayg, yeah, but human processes can have bugs, too. we *know* you can get into trouble, and we don't have a good way to fix it (yet; like i said, we're definitely working on it ;-) | 18:21 |
| clayg | 50+ GB databases SUCK | 18:21 |
| klamath_atx | thank you @timburke, gave me some info to digest. | 18:26 |
| *** rcernin has joined #openstack-swift | 20:33 | |
| *** rcernin has quit IRC | 20:37 | |
| *** rcernin has joined #openstack-swift | 21:02 | |
| openstackgerrit | Clay Gerrard proposed openstack/swift master: Enforce response length harder https://review.opendev.org/763405 | 21:10 |
| openstackgerrit | Tim Burke proposed openstack/python-swiftclient master: Allow tempurl times to have units https://review.opendev.org/758500 | 21:28 |
| *** jv has quit IRC | 21:33 | |
| *** jv has joined #openstack-swift | 21:44 | |
| *** dsariel has joined #openstack-swift | 21:58 | |
| *** dsariel has quit IRC | 22:03 | |
| *** baojg has quit IRC | 22:56 | |
| *** baojg has joined #openstack-swift | 22:57 | |
| *** baojg has quit IRC | 23:20 | |
| *** baojg has joined #openstack-swift | 23:21 | |
| *** rcernin has quit IRC | 23:35 | |
| *** rcernin has joined #openstack-swift | 23:35 | |
| *** baojg has quit IRC | 23:42 | |
| *** baojg has joined #openstack-swift | 23:42 | |
Generated by irclog2html.py 2.17.2 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!