Monday, 2018-10-29

« Sunday, 2018-10-28 Index Tuesday, 2018-10-30 »
kota_morning02:04
kota_big surprise with the REDHAT, IBM news.02:05
*** dhellmann has quit IRC02:55
*** dhellmann has joined #openstack-swift03:01
*** charz has joined #openstack-swift03:13
*** ccamacho has joined #openstack-swift07:07
*** gkadam has joined #openstack-swift07:44
*** pcaruana has joined #openstack-swift07:46
*** hseipp has joined #openstack-swift08:04
*** pvradu has joined #openstack-swift08:04
*** pvradu has quit IRC08:06
*** pvradu has joined #openstack-swift08:06
*** psachin has joined #openstack-swift08:07
*** gkadam has quit IRC08:22
*** gkadam has joined #openstack-swift08:22
*** e0ne has joined #openstack-swift08:45
*** gkadam has quit IRC08:53
*** gkadam has joined #openstack-swift08:55
*** mikecmpbll has joined #openstack-swift09:03
cschwedenotmyname: thx a lot! interesting times for sure09:18
openstackgerritRomain LE DISEZ proposed openstack/swift master: Make log format for requests configurable  https://review.openstack.org/54894809:51
*** ondrejme has joined #openstack-swift10:00
ondrejmeHi, I am running tempest on Ocata and in the test "tempest.api.object_storage.test_object_temp_url_negative.ObjectTempUrlNegativeTest.test_get_object_after_expiration_time" i have a problem that I am able to get an object through a temporary URL even after the URL expires. Any thoughts?10:02
*** hseipp has quit IRC10:02
*** ondrejme has quit IRC10:33
*** spsurya has quit IRC11:16
*** cloudnull has quit IRC11:16
*** zigo has quit IRC11:17
*** spsurya has joined #openstack-swift11:18
*** cloudnull has joined #openstack-swift11:23
*** e0ne has quit IRC11:25
*** e0ne has joined #openstack-swift11:34
*** ccamacho has quit IRC11:44
*** ccamacho has joined #openstack-swift11:45
*** StevenK has quit IRC12:32
*** StevenK has joined #openstack-swift12:32
*** e0ne has quit IRC12:58
seongsoocho Hi all~~ I need everyone's help!!  I have a terrible problem with put performance..13:10
seongsoochoHere is a detail of problem13:10
seongsoochohttp://lists.openstack.org/pipermail/openstack/2018-October/047109.html13:10
*** zigo has joined #openstack-swift13:12
*** e0ne has joined #openstack-swift13:19
*** e0ne has quit IRC13:24
*** e0ne has joined #openstack-swift13:30
*** hseipp has joined #openstack-swift13:49
*** mikecmpbll has quit IRC14:27
*** hseipp has quit IRC14:42
*** psachin has quit IRC14:43
*** psachin has joined #openstack-swift14:56
*** pvradu has quit IRC15:06
*** psachin has quit IRC15:07
*** itlinux has quit IRC15:09
*** irclogbot_2 has joined #openstack-swift15:11
*** admin6 has joined #openstack-swift15:20
*** SkyRocknRoll has joined #openstack-swift15:40
*** gyee has joined #openstack-swift15:40
*** hseipp has joined #openstack-swift15:55
*** itlinux has joined #openstack-swift16:01
*** SkyRocknRoll has quit IRC16:18
notmynamegood morning16:21
openstackgerritRomain LE DISEZ proposed openstack/swift master: Make log format for requests configurable  https://review.openstack.org/54894816:22
claygseongsoocho: you could decrease the node_timeout and let the proxy try to write more to handoffs16:34
claygseongsoocho: or you could try to use some of the ionice options (or other tuning options) to make the replicators hammer the disks a little less16:35
*** e0ne has quit IRC16:36
claygseongsoocho: could also be a io scheduling issue - are you using noop/deadline or cfq?  What kernel version?16:38
claygnotmyname: so I found myself thinking about the container-sharder (and db daemons in general) because of everyone seeming to be mad at them for pegging a core when they're not IO bound...16:43
claygthe object-updater and object-auditor both use ratelimit sleep to do a "X per second" sort of thing16:45
claygit would be pretty trivial to add that option to all the db deamons but thorwing a `RateLimitedIterator` around `roundrobin_datadirs`16:45
openstackgerritRomain LE DISEZ proposed openstack/swift master: SSYNC: Remove useless self.failures in sender  https://review.openstack.org/61398316:46
openstackgerritRomain LE DISEZ proposed openstack/swift master: SSYNC: Stop sharing a global connection  https://review.openstack.org/61398416:46
openstackgerritRomain LE DISEZ proposed openstack/swift master: SSYNC: Stop sharing a global response  https://review.openstack.org/61398516:46
openstackgerritRomain LE DISEZ proposed openstack/swift master: SSYNC: stop sharing global available_map/send_map  https://review.openstack.org/61398616:46
openstackgerritRomain LE DISEZ proposed openstack/swift master: SSYNC: enable multiple SSYNC connections per parts  https://review.openstack.org/61398716:46
claygmultiple SSYNC per part!?  *mind blown*16:46
rledisezclayg, the first 4 commits are just preparing the ground by moving some code around (not changing anything), the last one is to open mutiple connection to paralleize the jobs (x2.2 in performance in my tests)16:47
claygnotmyname: timburke: anyway, trying to decide if these cheap/obvious thing is best - or if we should looking further ahead and try to make the tunable option "I want my cycle time to be X hrs, go as fast as you need for that"16:47
claygrledisez: heyoh!  2x speedup ain't nothing!16:47
rlediseznext commit will by doing one syncfs() from time to time (x3.6 i tests), but I'll need some discussion with you about that :)16:48
claygwow!16:48
*** gkadam has quit IRC16:52
notmynameclayg: nah, I think the cheap/obvious thing is better than trying to predict what people will want in a tunable. and the first option doesn't preclude doing the other way later. I like the idea of "do a cycle in X time", but maybe that ends up being a more holistic change for a lot of daemons?16:52
*** mikecmpbll has joined #openstack-swift17:00
*** hseipp has quit IRC17:26
*** mikecmpbll has quit IRC17:38
*** mvkr has quit IRC18:05
*** mikecmpbll has joined #openstack-swift18:47
timburkeclayg: if you get to writing the patch for it, there's a bug you can attach: https://bugs.launchpad.net/swift/+bug/178475318:50
openstackLaunchpad bug 1784753 in OpenStack Object Storage (swift) "Add ratelimiting to more background daemons" [Undecided,New]18:50
claygtimburke: cool!18:50
*** mikecmpbll has quit IRC19:36
openstackgerritClay Gerrard proposed openstack/swift master: Add max_databases_per_second to db daemons  https://review.openstack.org/61401819:53
openstackgerritClay Gerrard proposed openstack/swift master: Add max_databases_per_second to db daemons  https://review.openstack.org/61401819:55
claygoh, i forgot to update configs and stuff20:08
openstackgerritClay Gerrard proposed openstack/swift master: Add max_databases_per_second to db daemons  https://review.openstack.org/61401820:37
*** mvkr has joined #openstack-swift21:21
*** e0ne has joined #openstack-swift21:29
*** e0ne has quit IRC21:36
*** itlinux has quit IRC21:46
*** pcaruana has quit IRC21:57
openstackgerritTim Burke proposed openstack/swift master: Add max_databases_per_second to db daemons  https://review.openstack.org/61401822:36
*** mikecmpbll has joined #openstack-swift23:07
timburkeanybody want to give https://review.openstack.org/#/c/611172/ a second look before I just +A? zaitcev's already got one +2, and there are a couple other approved patches that chain on it...23:28
patchbotpatch 611172 - swift - Unify handle_get/handle_head in decrypter - 1 patch set23:28
mattoliveraumorning23:28
mattoliverautimburke: I just got out of a meeting, let me grab coffee etc and then I'll review it and "hopefully" +A. Sorry I never got to it before.. been a little distracted this last week.23:30
timburkethanks! no worries -- i just wanted to help keep it moving :-)23:30
*** itlinux has joined #openstack-swift23:37
« Sunday, 2018-10-28 Index Tuesday, 2018-10-30 »

Generated by irclog2html.py 2.15.3 by Marius Gedminas - find it at mg.pov.lt!