| *** renich_ has quit IRC | 00:12 | |
| *** renich has joined #openstack-swift | 00:13 | |
| *** renich_ has joined #openstack-swift | 01:31 | |
| *** renich has quit IRC | 01:31 | |
| *** ccamacho has quit IRC | 02:50 | |
| *** psachin has joined #openstack-swift | 03:03 | |
| kota_ | hello world | 03:42 |
|---|---|---|
| mattoliverau | kota_: o/ | 03:59 |
| *** psachin has quit IRC | 04:30 | |
| mattoliverau | kota_: added this bug to zuul: https://storyboard.openstack.org/#!/story/2005238 They may mark it an invalid, but at least it's somewhere they'll see and they can ping me if they have questions :) | 04:50 |
| kota_ | mattoliverau: thanks a lot! | 04:51 |
| kota_ | you're master of storyboard. | 04:51 |
| mattoliverau | kota_: not really, just clicked buttons :P | 05:06 |
| kota_ | :P | 05:06 |
| *** rcernin has quit IRC | 05:58 | |
| *** rcernin has joined #openstack-swift | 06:13 | |
| *** e0ne has joined #openstack-swift | 06:24 | |
| *** e0ne has quit IRC | 06:27 | |
| *** hseipp has joined #openstack-swift | 07:27 | |
| *** renich_ has quit IRC | 07:34 | |
| *** pcaruana has joined #openstack-swift | 07:37 | |
| *** renich_ has joined #openstack-swift | 07:37 | |
| *** renich_ is now known as renich | 07:37 | |
| *** rcernin has quit IRC | 07:59 | |
| *** tkajinam has quit IRC | 08:18 | |
| *** pcaruana has quit IRC | 08:24 | |
| *** pcaruana|afk| has joined #openstack-swift | 08:24 | |
| *** renich has quit IRC | 08:24 | |
| *** e0ne has joined #openstack-swift | 08:35 | |
| *** ccamacho has joined #openstack-swift | 08:48 | |
| *** ccamacho has quit IRC | 08:52 | |
| *** e0ne has quit IRC | 11:03 | |
| *** e0ne has joined #openstack-swift | 12:12 | |
| *** FlorianFa has quit IRC | 12:14 | |
| DHE | has anyone tried bcache (or another SSD-offload/cache system) with swift? | 12:17 |
| *** pcaruana has joined #openstack-swift | 12:27 | |
| *** pcaruana|afk| has quit IRC | 12:28 | |
| *** henriqueof has joined #openstack-swift | 12:49 | |
| *** pcaruana has quit IRC | 13:36 | |
| *** e0ne has quit IRC | 13:58 | |
| *** e0ne has joined #openstack-swift | 14:06 | |
| *** pcaruana has joined #openstack-swift | 14:17 | |
| *** e0ne has quit IRC | 15:51 | |
| *** renich has joined #openstack-swift | 15:53 | |
| *** renich has quit IRC | 16:05 | |
| *** renich_ has joined #openstack-swift | 16:05 | |
| notmyname | good morning | 16:20 |
| notmyname | DHE: good question (bcache). I seem to remember someone mentioning trying it, but not being too happy with the results. TBH I don't remember more than that | 16:21 |
| DHE | ZFS has a rather cool feature in the pre-release version where metadata (and only metadata) can be offloaded to an SSD. that would make replication largely laugh off any disk IO work but put the bulk of the file storage on the spinning disks | 16:23 |
| *** gyee has joined #openstack-swift | 16:23 | |
| notmyname | it is a cool feature. but is that only in a pre-release. I thought I'd seen something like that in the past. index offloading, or something like that | 16:26 |
| notmyname | a few years ago I played with a pretty cool library from Intel that does something similar for any file system, and we saw *great* performance improvements in swift with it | 16:27 |
| DHE | it's ZFS-specific. and the feature has been generally available for a year or two for people willing to build from a github Pull Request. It was explicitly merged into the main Git branch a few months ago | 16:27 |
| notmyname | however, either the ZFS thing or the Intel CAS library will both add operational complexity | 16:27 |
| DHE | that's the Optane thing? | 16:28 |
| *** _david_sohonet has joined #openstack-swift | 16:28 | |
| notmyname | hmm... I must be remembering something different in ZFS. whatever I remember was from back in the solaris/opensolaris days | 16:28 |
| notmyname | let me find something on CAS | 16:28 |
| notmyname | and CAS isnt' (wasn't?) restricted to optane. although they want you to buy optane, of course :-) | 16:28 |
| *** ybunker has joined #openstack-swift | 16:29 | |
| _david_sohonet | hi | 16:29 |
| notmyname | DHE: https://www.intel.com/content/www/us/en/software/intel-cache-acceleration-software-performance.html | 16:29 |
| DHE | historically ZFS has had something called L2ARC, which basically trickles RAM cache onto an SSD. unfortunately this doesn't survive a reboot, and it's a pretty dumb cache system. eg: it just writes data from start to end on the SSD, then it starts again, overwriting data it previously wrote with no regard for popularity | 16:29 |
| _david_sohonet | I logged issue https://bugs.launchpad.net/swift/+bug/1819252 a while ago. Is there someone that will triage new issues as they arrive ? | 16:31 |
| openstack | Launchpad bug 1819252 in OpenStack Object Storage (swift) "connection closed before XML bulk delete data can be transmitted" [Undecided,New] | 16:31 |
| notmyname | the CAS library is able to detect a filesystem write and cache it if it's in a particular class of data. either based on size or if it's filesystem metadata. let me try to find some better links | 16:31 |
| ybunker | hi all, quick questions... is possible to upgrade keystone (juno) release to rocky release without any further outages? | 16:33 |
| DHE | that's quite a version jump. Sounds like you might have an extended outage there... | 16:35 |
| notmyname | ybunker: that's a question for the keystone team. but from the swift perspective, you can upgrade from any older version of swift to the current version without any end-user downtime (normally, depending on your cluster config) | 16:35 |
| ybunker | notmyname: oh ok, thanks a lot :), i will send the questions to the keystone team | 16:36 |
| notmyname | DHE: https://learn.swiftstack.com/hubfs/mdoc/SwiftStack_Solution_Brief_Final.pdf and https://learn.swiftstack.com/hubfs/mdoc/SwiftStack_CAS_White_Paper_Final.pdf | 16:41 |
| notmyname | DHE: I don't seem to have enough google-fu this morning to find a video. I'm pretty sure we did a talk, but I don't remember | 16:41 |
| notmyname | _david_sohonet: good morning | 16:41 |
| notmyname | or likely afternoon for you if you're with sohonet :-) | 16:42 |
| notmyname | _david_sohonet: off the top of my head, the best person who could answer that is timburke, but he's taking yesterday and today off. | 16:43 |
| *** renich_ has quit IRC | 16:44 | |
| _david_sohonet | Gotcha, thanks notmyname | 16:45 |
| notmyname | _david_sohonet: but this does look like something that needs to be triaged. the bug report looks good; thanks for leaving a good one :-) | 16:45 |
| *** renich has joined #openstack-swift | 16:46 | |
| *** henriqueof has quit IRC | 16:48 | |
| *** ybunker has quit IRC | 16:50 | |
| *** renich has quit IRC | 16:50 | |
| DHE | notmyname: cool... | 16:51 |
| *** renich has joined #openstack-swift | 16:52 | |
| *** e0ne has joined #openstack-swift | 17:10 | |
| *** e0ne has quit IRC | 17:11 | |
| notmyname | timburke: clayg: mattoliverau: I think https://bugs.launchpad.net/swift/+bug/1819252 looks like a regression | 17:11 |
| openstack | Launchpad bug 1819252 in OpenStack Object Storage (swift) "connection closed before XML bulk delete data can be transmitted" [Undecided,New] | 17:11 |
| notmyname | _david_sohonet: the heartbeating functionality that patch 510715 fixed is important. so I don't think we can or should simply revert the patch | 17:12 |
| patchbot | https://review.openstack.org/#/c/510715/ - swift - Fix bulk heartbeating when emitting XML (MERGED) - 1 patch set | 17:12 |
| notmyname | warrants some further investigation | 17:12 |
| *** ybunker has joined #openstack-swift | 17:13 | |
| *** itlinux has quit IRC | 17:18 | |
| *** ybunker has quit IRC | 17:37 | |
| *** ybunker has joined #openstack-swift | 17:38 | |
| notmyname | _david_sohonet: I've confirmed the bug on my saio | 17:46 |
| notmyname | timburke: tdasilva: clayg: would be great to find a fix for https://bugs.launchpad.net/swift/+bug/1819252 and land it before we tag a release | 17:48 |
| openstack | Launchpad bug 1819252 in OpenStack Object Storage (swift) "connection closed before XML bulk delete data can be transmitted" [Undecided,Confirmed] | 17:48 |
| notmyname | to which you're thinking, "hmm... sure would be. don't you have a text editor, notmyname?" ;-) | 17:48 |
| _david_sohonet | lol :-) | 17:50 |
| notmyname | want to hear a funny story? we don't have any bulk unit tests that use `Accept: application/xml` and we don't have any functests at all for bulk | 18:17 |
| notmyname | hilarious! | 18:17 |
| *** hseipp has quit IRC | 18:17 | |
| notmyname | oh, it's text/xml, and we do test that | 18:18 |
| _david_sohonet | I was wondering how this got through testing .... | 18:18 |
| *** e0ne has joined #openstack-swift | 18:53 | |
| notmyname | _david_sohonet: oh! I'm getting a slightly different error than you are | 19:06 |
| notmyname | I updated launchpad | 19:11 |
| _david_sohonet | We are running 2.17 Swift, I did not test a more recent release | 19:29 |
| *** e0ne has quit IRC | 19:37 | |
| *** openstack has joined #openstack-swift | 19:43 | |
| *** ChanServ sets mode: +o openstack | 19:43 | |
| *** pcaruana has quit IRC | 19:59 | |
| notmyname | _david_sohonet: I think I have a patch | 22:00 |
| notmyname | local curl tests passed. running in-repo tests. still need to think if it's possible to write some sort of other test for it | 22:00 |
| _david_sohonet | Sounds good, thanks notmyname . It seems you are still handy with the text editor :-) | 22:03 |
| notmyname | no one is as shocked as I am! | 22:04 |
| openstackgerrit | John Dickinson proposed openstack/swift master: Fix bulk responses when using xml and Expect 100-continue https://review.openstack.org/643684 | 22:28 |
| openstackgerrit | John Dickinson proposed openstack/swift master: Fix bulk responses when using xml and Expect 100-continue https://review.openstack.org/643684 | 22:28 |
| notmyname | clayg: timburke: ^^ I'd appreciate your eyes on that patch | 22:30 |
| notmyname | _david_sohonet: I'm not suggesting that you apply that patch to your prod cluster, but if you did, it would probably fix your issue | 22:31 |
| openstackgerrit | John Dickinson proposed openstack/swift master: Fix bulk responses when using xml and Expect 100-continue https://review.openstack.org/643684 | 22:56 |
| notmyname | now with a functional test! | 22:56 |
| notmyname | bah hang on. test doesn't fail on master | 23:02 |
| notmyname | wait | 23:02 |
| notmyname | I didn't run it | 23:02 |
| notmyname | ok, yeah, it's good. fails on master, works on the patch. actually tests the bug that _david_sohonet discovered | 23:03 |
| notmyname | I'd still like eyes from clayg and/or timburke on it :-) | 23:03 |
Generated by irclog2html.py 2.15.3 by Marius Gedminas - find it at mg.pov.lt!