| *** efried has quit IRC | 00:48 | |
| *** efried has joined #openstack-swift | 00:49 | |
| openstackgerrit | Merged openstack/swift stable/stein: Update known-failures and config for up-rev'ed ceph/s3tests https://review.opendev.org/697350 | 02:20 |
|---|---|---|
| *** gyee has quit IRC | 02:30 | |
| kota_ | morning | 03:08 |
| *** psachin has joined #openstack-swift | 03:41 | |
| *** openstackstatus has joined #openstack-swift | 03:42 | |
| *** ChanServ sets mode: +v openstackstatus | 03:42 | |
| zaitcev | Guys, does anyone have a small production cluster, where you could run this on one of nodes (I mean, without causing harm): find /srv/node -name '*.db'| xargs stat -t | awk '{print $2, $1}' | sort -n -r | head -20 | 04:00 |
| zaitcev | My longest is only 77MB | 04:00 |
| sorrison | currently running this on a nectar swift node (been running for a few mins so far) | 04:10 |
| sorrison | 580444160 is the biggest | 04:12 |
| sorrison | this in bytes? | 04:12 |
| sorrison | 554M | 04:14 |
| *** tkajinam has quit IRC | 05:02 | |
| *** tkajinam has joined #openstack-swift | 05:04 | |
| *** tkajinam_ has joined #openstack-swift | 05:35 | |
| *** tkajinam has quit IRC | 05:38 | |
| *** rcernin has quit IRC | 06:03 | |
| zaitcev | That sounds reasonable. | 06:12 |
| *** diablo_rojo has quit IRC | 07:23 | |
| *** rcernin has joined #openstack-swift | 07:25 | |
| *** pcaruana has joined #openstack-swift | 08:11 | |
| *** tkajinam_ has quit IRC | 08:16 | |
| *** rpittau|afk is now known as rpittau | 08:21 | |
| *** rdejoux has joined #openstack-swift | 08:22 | |
| *** tesseract has joined #openstack-swift | 08:31 | |
| *** rcernin has quit IRC | 10:12 | |
| *** rcernin has joined #openstack-swift | 10:16 | |
| *** rcernin has quit IRC | 10:33 | |
| openstackgerrit | Romain LE DISEZ proposed openstack/swift master: Replace all "with Chunk*Timeout" by a watchdog https://review.opendev.org/697653 | 11:53 |
| *** rdejoux has quit IRC | 13:38 | |
| openstackgerrit | Romain LE DISEZ proposed openstack/swift master: Replace all "with Chunk*Timeout" by a watchdog https://review.opendev.org/697653 | 14:01 |
| DHE | I have an 11 GB database representing a container with ~47 million items in it | 14:06 |
| clayg | 47M is pretty big! 👍 | 14:41 |
| clayg | DHE: you gunna try and shard it? | 14:41 |
| DHE | considering it since 60 million is the short term goal (still loading data into it) | 14:42 |
| DHE | once that's done I will need to figure out which buttons to press to shard the container | 15:02 |
| clayg | https://docs.openstack.org/swift/latest/overview_container_sharding.html | 15:05 |
| clayg | yeah it's a party! | 15:05 |
| zaitcev | Holy cow, 47 million | 15:10 |
| DHE | I just said I ain't done yet! | 15:10 |
| DHE | :) | 15:11 |
| openstackgerrit | Romain LE DISEZ proposed openstack/swift master: Replace all "with Chunk*Timeout" by a watchdog https://review.opendev.org/697653 | 15:11 |
| zaitcev | So, it's only 237 bytes per object. That's quite efficient. | 15:16 |
| zaitcev | I was going to ask you if you could vacuum it, but I see now that it's pointless. | 15:17 |
| DHE | seems reasonable for sqlite | 15:49 |
| timburke | good morning | 16:21 |
| *** efried is now known as efried_afk | 16:22 | |
| openstackgerrit | Tim Burke proposed openstack/swift master: Set swift_source more in s3api middleware https://review.opendev.org/697580 | 16:32 |
| timburke | well that's a little weird... https://zuul.opendev.org/t/openstack/build/95160b1006c44caca21f20515e5ff1e2/console#7/1/0/proxy1 | 16:37 |
| timburke | rc of -13 -- i guess ansible got a SIGPIPE? | 16:37 |
| *** gyee has joined #openstack-swift | 16:48 | |
| *** rpittau is now known as rpittau|afk | 16:51 | |
| *** psachin has quit IRC | 16:55 | |
| *** efried_afk is now known as efried | 17:00 | |
| *** pcaruana has quit IRC | 17:23 | |
| *** tesseract has quit IRC | 17:35 | |
| clayg | timburke: so yeah container-sync doesn't work on versioned containers... | 18:29 |
| openstackgerrit | Romain LE DISEZ proposed openstack/swift master: Replace all "with Chunk*Timeout" by a watchdog https://review.opendev.org/697653 | 18:30 |
| timburke | clayg, i'm not too surprised :-/ | 18:30 |
| timburke | i guess the question is: how *should* it work? | 18:31 |
| clayg | i don't think i really had an opinion, the traceback is just "unkown exception" in container_sync_row 🤷♂️ | 18:31 |
| timburke | from the 409 on PUT, i'm guessing? | 18:32 |
| clayg | "Unknown exception trying to GET" | 18:33 |
| timburke | oh... hmm... curious | 18:34 |
| clayg | so after fixing a weird bug with x-timestamp vs x-backend-timestamp the GET works... | 19:00 |
| clayg | basically we're trying to sync a static-link and the remote PUT 412's because there's no matching target 🤷♂️ | 19:01 |
| timburke | mm... 412 because the target's in the (inaccesible) reserved namespace... | 19:03 |
| openstackgerrit | Tim Burke proposed openstack/swift master: s3api: Implement object versioning API https://review.opendev.org/673682 | 19:13 |
| timburke | ^^^ should make s3api happy again when object_versioning is disabled | 19:13 |
| openstackgerrit | Merged openstack/swift stable/train: Update known-failures and config for up-rev'ed ceph/s3tests https://review.opendev.org/697349 | 19:24 |
| openstackgerrit | Merged openstack/swift master: Add option for debug query logging https://review.opendev.org/691399 | 19:54 |
| clayg | timburke: so I guess container-sync never "works" with static links... unless you somehow configure the targets to sync also? | 20:33 |
| clayg | well, i guess that's true for symlinks too - it might "work" - but it wouldn't be useful unless you've also synced the targets | 20:34 |
| clayg | I wonder if adding symlink=get to container sync was a mistake? | 20:34 |
| timburke | check: https://github.com/openstack/swift/blob/master/test/probe/test_container_sync.py#L446-L464 | 20:35 |
| timburke | auth becomes a problem w/o symlink=get | 20:35 |
| timburke | go create some symlinks pointing to names you don't have access to (but suspect exist), wait for container sync to exfiltrate someone else's data for you | 20:37 |
| *** diablo_rojo has joined #openstack-swift | 20:57 | |
| clayg | SO GOOD CALL WITH THE symlin=get THEN! | 21:00 |
| clayg | anyway, it's not obvious to me what a fix would look like - i guess I'll leave this one for now, work on other stuff and keep thinking about it | 21:00 |
| timburke | i *think* it might be reasonable (or at any rate, both defensible and implementable) to have container sync be smart enough to see that it was a versioning symlink and follow it manually | 21:10 |
| *** irclogbot_3 has quit IRC | 21:45 | |
| *** diablo_rojo has quit IRC | 22:31 | |
| openstackgerrit | Tim Burke proposed openstack/swift master: WIP: Have slo tell the object-server that it wants whole manifests https://review.opendev.org/697739 | 23:27 |
Generated by irclog2html.py 2.15.3 by Marius Gedminas - find it at mg.pov.lt!