*** jerrygb_ has joined #openstack-swift | 00:05 | |
*** jerrygb_ has quit IRC | 00:08 | |
*** minwoob has quit IRC | 00:09 | |
*** jerrygb has quit IRC | 00:09 | |
*** mahatic has joined #openstack-swift | 00:26 | |
*** garthb has quit IRC | 00:30 | |
*** mahatic has quit IRC | 00:37 | |
*** mfalatic has quit IRC | 00:40 | |
*** gyee has quit IRC | 00:49 | |
*** peluse has quit IRC | 00:57 | |
*** km has quit IRC | 01:03 | |
*** km has joined #openstack-swift | 01:04 | |
*** DericHorn-HP has joined #openstack-swift | 01:07 | |
*** peluse has joined #openstack-swift | 01:15 | |
*** DericHorn-HP has quit IRC | 01:17 | |
*** DericHorn-HP has joined #openstack-swift | 01:27 | |
*** rohit_ has quit IRC | 01:40 | |
*** jerrygb has joined #openstack-swift | 01:44 | |
*** jerrygb has quit IRC | 01:45 | |
*** jerrygb has joined #openstack-swift | 01:45 | |
*** haomaiwang has joined #openstack-swift | 01:58 | |
*** haomaiwang has quit IRC | 02:01 | |
*** haomaiwang has joined #openstack-swift | 02:01 | |
*** DericHorn-HP has quit IRC | 02:04 | |
*** DericHorn-HP has joined #openstack-swift | 02:05 | |
*** DericHorn-HP has quit IRC | 02:06 | |
*** DericHorn-HP has joined #openstack-swift | 02:07 | |
wbhuber | clayg: u in? | 02:16 |
---|---|---|
*** darrenc is now known as darrenc_afk | 02:21 | |
*** DericHorn-HP has quit IRC | 02:34 | |
*** DericHorn-HP has joined #openstack-swift | 02:37 | |
*** early has quit IRC | 02:47 | |
mattoliverau | wbhuber: haven't seen him since I've been up. But he may pop up cause he keeps some wierd times :) | 02:47 |
wbhuber | mattoliverau: thanx. | 02:48 |
*** darrenc_afk is now known as darrenc | 02:49 | |
*** sanchitmalhotra has joined #openstack-swift | 02:52 | |
*** early has joined #openstack-swift | 02:55 | |
*** wbhuber has quit IRC | 02:56 | |
*** haomaiwang has quit IRC | 03:01 | |
*** haomaiwa_ has joined #openstack-swift | 03:01 | |
*** DericHorn-HP has quit IRC | 03:04 | |
*** DericHorn-HP has joined #openstack-swift | 03:05 | |
*** david-lyle has joined #openstack-swift | 03:13 | |
*** peluse has quit IRC | 03:21 | |
notmyname | good evening | 03:25 |
openstackgerrit | Merged openstack/swift: Fix proxy handling of EC client disconnect https://review.openstack.org/225357 | 03:29 |
notmyname | yay | 03:29 |
notmyname | I'm working on a list of prioritized reviews tonight. that's one of them I can take off now :-) | 03:30 |
*** DericHorn-HP has quit IRC | 03:38 | |
mattoliverau | notmyname: welcome back :) | 03:40 |
*** DericHorn-HP has joined #openstack-swift | 03:41 | |
*** jrichli has joined #openstack-swift | 03:49 | |
notmyname | clayg: on your note on outstanding patches, you suggested that https://bugs.launchpad.net/swift/+bug/1484598 was potentially less that a critical bug for release (and patch https://review.openstack.org/#/c/213147/) | 03:54 |
openstack | Launchpad bug 1484598 in OpenStack Object Storage (swift) "Proxy server ignores additional fragments on primary nodes" [High,In progress] - Assigned to paul luse (paul-e-luse) | 03:54 |
notmyname | why would it be anything other than critical for the release? seems pretty important | 03:54 |
*** hrou has quit IRC | 03:56 | |
*** jerrygb has quit IRC | 03:56 | |
*** mahatic has joined #openstack-swift | 03:56 | |
*** garthb has joined #openstack-swift | 03:57 | |
mahatic | good morning | 03:59 |
*** DericHorn-HP has quit IRC | 03:59 | |
*** haomaiwa_ has quit IRC | 04:01 | |
*** haomaiwang has joined #openstack-swift | 04:01 | |
*** silor has joined #openstack-swift | 04:08 | |
mattoliverau | mahatic: morning :) | 04:09 |
notmyname | https://wiki.openstack.org/wiki/Swift/PriorityReviews is updated with a "Liberty patches" section. it's got a bunch of EC stuff in it right now | 04:10 |
notmyname | C = critical, H = high, L = low | 04:10 |
notmyname | for priority | 04:10 |
notmyname | the critical and high ones are starred in the review dashboard | 04:11 |
mahatic | mattoliverau: hello | 04:12 |
notmyname | also, this will be the focus of the meeting this week | 04:12 |
zaitcev | hmm, 215276 is WIP | 04:13 |
mattoliverau | Kk | 04:13 |
notmyname | zaitcev: ah, thanks. I'll ad that note | 04:13 |
*** jrichli has quit IRC | 04:14 | |
notmyname | yeah, we need clayg to figure out what's going on and what's needed | 04:14 |
notmyname | the other big deal (related, though not directly, to EC) is https://bugs.launchpad.net/swift/+bug/1452431 | 04:17 |
openstack | Launchpad bug 1452431 in OpenStack Object Storage (swift) "some parts replicas assigned to duplicate devices in the ring" [Critical,Confirmed] - Assigned to Samuel Merritt (torgomatic) | 04:17 |
notmyname | ah, wait. that patch 222799 is already referenced :-) | 04:17 |
patchbot | notmyname: https://review.openstack.org/#/c/222799/ | 04:17 |
notmyname | well, if we can get that patch, then we can likely have a little bit of breathing room for really solving the ring placement bug | 04:18 |
notmyname | clayg: link to the error in the ring validator patch for py27 tests in the gate: http://logs.openstack.org/99/222799/4/check/gate-swift-python27/a8ca52c/console.html#_2015-09-22_23_40_46_866 | 04:19 |
notmyname | acoles_: clayg: tomorrow at the meeting here's what I'd love to see: | 04:21 |
*** zaitcev has quit IRC | 04:22 | |
notmyname | we've got a good list of patches and bug that need closing. but getting a bigger picture of it would be helpful. for me, especially the multiple versions of a FA and what to do in that case would be nice to know | 04:23 |
notmyname | there are a few patches and comments in gerrit that leave me wondering which way the current plan is for | 04:23 |
notmyname | ie for optimistic gets | 04:25 |
*** silor1 has joined #openstack-swift | 04:25 | |
* mahatic will be right back | 04:26 | |
*** silor has quit IRC | 04:26 | |
*** silor1 is now known as silor | 04:26 | |
*** mahatic has quit IRC | 04:26 | |
clayg | notmyname: the duplicate parts on one node issue takes two things 1) a part on a handoff 2) a rebalance causing that part to be a primary - after minwoo's patch 213147 we'll have tests that validate the reconstructor can dig it self out of this situation, and if it can keep up with the ring's min_part_hours we won't ever have more than a single unavailable frag while it does - meaning there's no risk to availability and dura | 04:47 |
patchbot | clayg: https://review.openstack.org/#/c/213147/ | 04:47 |
notmyname | clayg: ack | 04:50 |
clayg | notmyname: so I don't think lp bug #1484598 is really much of an issue now that we have etag buckets and use handoffs correctly to work around overwrite | 04:50 |
openstack | Launchpad bug 1484598 in OpenStack Object Storage (swift) "Proxy server ignores additional fragments on primary nodes" [High,In progress] https://launchpad.net/bugs/1484598 - Assigned to paul luse (paul-e-luse) | 04:50 |
clayg | notmyname: i'm cool improving on the proxy behavior - but it should fall out of the work on optomistic GET's - where we're evaluating different mixtures of available data and potentially making mutltiple requests to a single node *already* - rather than be an independent behavior | 04:51 |
notmyname | clayg: and because of the etag buckets and handoffs on overwrites, that's why the stuff you and acoles_ have been doing around the optimistic gets is also listed as WIP--basically, not sure yet if it's needed. patch 181407 and patch 174875 | 04:53 |
patchbot | notmyname: https://review.openstack.org/#/c/181407/ | 04:53 |
patchbot | notmyname: https://review.openstack.org/#/c/174875/ | 04:53 |
notmyname | and maybe even patch 215276 | 04:54 |
patchbot | notmyname: https://review.openstack.org/#/c/215276/ | 04:54 |
clayg | notmyname: well, there was another one that's already landed - paul started it - ended up splitting node_iter into a NodeIter class - it helped a lot | 04:54 |
clayg | ... but it merged awhile ago (shortly after Austin) | 04:54 |
*** garthb has quit IRC | 04:55 | |
clayg | notmyname: i briefly talked to acoles_ about patch 181407 and patch 174875 - they're not critical - some of the bad behaviors they were trying to address have been addressed in subsequent patches (we had better ideas) - so they need to be compleatly re-evaluated and there's no open bugs related to the issues they set out to address currently open (issues they were trying to expose have been apparently mostly fixed in other w | 04:56 |
patchbot | clayg: https://review.openstack.org/#/c/181407/ | 04:56 |
patchbot | clayg: https://review.openstack.org/#/c/174875/ | 04:56 |
notmyname | ok | 04:57 |
notmyname | clayg: I took your gist and put it into the wiki at https://wiki.openstack.org/wiki/Swift/PriorityReviews | 04:57 |
notmyname | so we have a common place to look | 04:57 |
clayg | patch 215276 is a beast - thank god for acoles_ - but he's probably going to need some help and I'm frankly thinking that aside from maybe some cleanup on the wire-protocol-timeout-disconnect stuff I'll be mainly working on rings | 04:57 |
patchbot | clayg: https://review.openstack.org/#/c/215276/ | 04:57 |
clayg | yeah I saw that on the wiki page - looks nice | 04:57 |
clayg | ... but - I have an idea on the rings! | 04:58 |
notmyname | yay! | 04:58 |
clayg | well see - i'm going back to my debugger | 04:58 |
*** haomaiwang has quit IRC | 05:01 | |
*** haomaiwang has joined #openstack-swift | 05:01 | |
*** zhill has joined #openstack-swift | 05:03 | |
*** mahatic has joined #openstack-swift | 05:04 | |
*** klrmn1 has quit IRC | 05:05 | |
*** ppai has joined #openstack-swift | 05:07 | |
*** trifon has joined #openstack-swift | 05:18 | |
*** dipe has joined #openstack-swift | 05:27 | |
*** proteusguy__ has quit IRC | 05:33 | |
*** zhill has quit IRC | 05:40 | |
*** proteusguy__ has joined #openstack-swift | 05:45 | |
*** haomaiwang has quit IRC | 06:01 | |
*** haomaiwang has joined #openstack-swift | 06:01 | |
*** jerrygb has joined #openstack-swift | 06:08 | |
*** jerrygb has quit IRC | 06:12 | |
*** SkyRocknRoll has joined #openstack-swift | 06:12 | |
*** sanchitmalhotra1 has joined #openstack-swift | 06:29 | |
*** sanchitmalhotra has quit IRC | 06:31 | |
*** sanchitmalhotra has joined #openstack-swift | 06:45 | |
*** sanchitmalhotra1 has quit IRC | 06:47 | |
*** haomaiwang has quit IRC | 07:01 | |
*** haomaiwang has joined #openstack-swift | 07:01 | |
*** rledisez has joined #openstack-swift | 07:07 | |
*** mahatic has quit IRC | 07:11 | |
*** dipe has quit IRC | 07:33 | |
*** sayali has quit IRC | 07:34 | |
*** sayali has joined #openstack-swift | 07:38 | |
*** jordanP has joined #openstack-swift | 07:41 | |
*** KLuka_ has left #openstack-swift | 07:48 | |
*** geaaru has joined #openstack-swift | 07:51 | |
*** sanchitmalhotra1 has joined #openstack-swift | 07:57 | |
*** acoles_ is now known as acoles | 07:58 | |
*** sanchitmalhotra has quit IRC | 07:59 | |
openstackgerrit | Clay Gerrard proposed openstack/swift: Validate against duplicate part replica device assignment https://review.openstack.org/222799 | 08:00 |
openstackgerrit | Clay Gerrard proposed openstack/swift: wip: figure out when you're going to have a bad time https://review.openstack.org/226645 | 08:00 |
*** haomaiwang has quit IRC | 08:01 | |
*** haomaiwang has joined #openstack-swift | 08:01 | |
*** haomaiwang has quit IRC | 08:10 | |
*** haomaiwa_ has joined #openstack-swift | 08:12 | |
*** mahatic has joined #openstack-swift | 08:13 | |
*** aix has joined #openstack-swift | 08:15 | |
*** sanchitmalhotra has joined #openstack-swift | 08:17 | |
*** sanchitmalhotra1 has quit IRC | 08:20 | |
*** mac_ified has quit IRC | 08:27 | |
openstackgerrit | Merged openstack/swift: Change fake_footers to replace response headers instead of append https://review.openstack.org/222953 | 08:37 |
openstackgerrit | Merged openstack/swift: Minor keymaster cleanup https://review.openstack.org/226531 | 08:51 |
openstackgerrit | Alistair Coles proposed openstack/python-swiftclient: Make more assertions in client unit tests https://review.openstack.org/226654 | 08:57 |
acoles | timburke: see what you started ^^ :) | 08:57 |
*** sanchitmalhotra1 has joined #openstack-swift | 08:59 | |
*** haomaiwa_ has quit IRC | 09:01 | |
*** 16WAAB8UT has joined #openstack-swift | 09:01 | |
acoles | notmyname: i agree with clayg re patch 181407 and patch 174875. there are some optimisations in them but they need rework and imho not critical for liberty. | 09:02 |
patchbot | acoles: https://review.openstack.org/#/c/181407/ | 09:02 |
patchbot | acoles: https://review.openstack.org/#/c/174875/ | 09:02 |
*** sanchitmalhotra has quit IRC | 09:02 | |
acoles | clayg: thanks for review on 215276, yes its a beast. | 09:05 |
*** jerrygb has joined #openstack-swift | 09:08 | |
*** jerrygb has quit IRC | 09:13 | |
*** joeljwright has joined #openstack-swift | 09:18 | |
*** ChanServ sets mode: +v joeljwright | 09:18 | |
*** joeljwright has quit IRC | 09:19 | |
*** sanchitmalhotra has joined #openstack-swift | 09:27 | |
*** sanchitmalhotra1 has quit IRC | 09:30 | |
*** tsubic has quit IRC | 09:44 | |
*** tsubic has joined #openstack-swift | 09:44 | |
*** 16WAAB8UT has quit IRC | 10:01 | |
*** 6JTAB3MXI has joined #openstack-swift | 10:01 | |
*** Kennan_Vacation has quit IRC | 10:01 | |
*** Kennan_Vacation has joined #openstack-swift | 10:02 | |
*** km has quit IRC | 10:14 | |
*** ctennis has quit IRC | 10:41 | |
*** ctennis has joined #openstack-swift | 10:41 | |
*** mahatic has quit IRC | 10:47 | |
*** 6JTAB3MXI has quit IRC | 11:01 | |
*** haomaiwang has joined #openstack-swift | 11:01 | |
*** aix has quit IRC | 11:03 | |
openstackgerrit | Alistair Coles proposed openstack/python-swiftclient: Add tests and param definitions for headers parameter https://review.openstack.org/226236 | 11:09 |
*** kei_yama has quit IRC | 11:20 | |
openstackgerrit | Alistair Coles proposed openstack/python-swiftclient: Add tests and param definitions for headers parameter https://review.openstack.org/226236 | 11:22 |
*** haomaiwang has quit IRC | 11:22 | |
*** shakamunyi has quit IRC | 11:25 | |
*** aix has joined #openstack-swift | 11:34 | |
*** mahatic has joined #openstack-swift | 11:34 | |
*** jerrygb has joined #openstack-swift | 11:35 | |
*** ccavanna has quit IRC | 11:36 | |
openstackgerrit | Donagh McCabe proposed openstack/swift: Disallow cross-reseller COPY https://review.openstack.org/226707 | 11:38 |
*** jerrygb has quit IRC | 11:39 | |
*** sanchitmalhotra1 has joined #openstack-swift | 11:42 | |
*** sanchitmalhotra has quit IRC | 11:43 | |
*** jerrygb has joined #openstack-swift | 11:48 | |
*** sayali has quit IRC | 11:50 | |
*** sayali has joined #openstack-swift | 11:55 | |
tamizh_geek | clayg: Hello, when you have time can you please review this for me? https://review.openstack.org/#/c/225499/ | 11:58 |
*** sanchitmalhotra has joined #openstack-swift | 12:07 | |
*** sanchitmalhotra1 has quit IRC | 12:08 | |
*** agomez has joined #openstack-swift | 12:12 | |
*** mahatic has quit IRC | 12:21 | |
*** ccavanna has joined #openstack-swift | 12:24 | |
*** mahatic has joined #openstack-swift | 12:39 | |
*** agomez has quit IRC | 12:47 | |
*** sayali has quit IRC | 12:47 | |
*** sayali has joined #openstack-swift | 12:50 | |
*** ppai has quit IRC | 12:58 | |
*** peluse has joined #openstack-swift | 12:58 | |
*** janonymous has quit IRC | 13:00 | |
*** mac_ified has joined #openstack-swift | 13:01 | |
*** mac_ified has quit IRC | 13:01 | |
*** hrou has joined #openstack-swift | 13:01 | |
*** jkugel has joined #openstack-swift | 13:02 | |
*** haomaiwang has joined #openstack-swift | 13:08 | |
hrou | acoles, thanks for reviewing / merging that crypto patch, yea I'm willing to bet 12 hours is a record ! But careful now you're setting a precedence ; ) | 13:17 |
*** ccavanna has quit IRC | 13:19 | |
*** silor has quit IRC | 13:22 | |
*** wbhuber has joined #openstack-swift | 13:25 | |
*** dustins has joined #openstack-swift | 13:28 | |
*** SkyRocknRoll has quit IRC | 13:36 | |
*** sanchitmalhotra1 has joined #openstack-swift | 13:39 | |
*** ccavanna has joined #openstack-swift | 13:40 | |
*** sanchitmalhotra has quit IRC | 13:42 | |
*** mahatic has quit IRC | 13:44 | |
*** sanchitmalhotra has joined #openstack-swift | 13:53 | |
peluse | FYI the perf data I posted yesterday has some issues with the bar charts at the end - the 50/50 numbers are all about 50% of what they should be. Mistake in compiling the data, will post a new one later today | 13:56 |
*** delatte has quit IRC | 13:56 | |
*** sanchitmalhotra1 has quit IRC | 13:56 | |
*** delattec has quit IRC | 13:58 | |
*** kairo has quit IRC | 13:59 | |
*** kairo has joined #openstack-swift | 13:59 | |
*** haomaiwang has quit IRC | 14:01 | |
*** haomaiwang has joined #openstack-swift | 14:01 | |
*** sanchitmalhotra has quit IRC | 14:02 | |
*** jerrygb has quit IRC | 14:03 | |
*** trifon has quit IRC | 14:04 | |
*** jerrygb has joined #openstack-swift | 14:07 | |
*** breitz has quit IRC | 14:10 | |
*** breitz has joined #openstack-swift | 14:10 | |
openstackgerrit | OpenStack Proposal Bot proposed openstack/python-swiftclient: Updated from global requirements https://review.openstack.org/89250 | 14:12 |
openstackgerrit | OpenStack Proposal Bot proposed openstack/swift: Updated from global requirements https://review.openstack.org/88736 | 14:13 |
*** mahatic has joined #openstack-swift | 14:13 | |
*** jlhinson has joined #openstack-swift | 14:17 | |
*** hemanthm has left #openstack-swift | 14:18 | |
*** jrichli has joined #openstack-swift | 14:19 | |
*** jkugel has left #openstack-swift | 14:23 | |
*** pberis has joined #openstack-swift | 14:31 | |
*** cdelatte has joined #openstack-swift | 14:36 | |
*** delattec has joined #openstack-swift | 14:36 | |
clayg | acoles: i hope you mean that thank you - it was just my first impression - you know this one -> http://www.osnews.com/story/19266/WTFs_m | 14:42 |
clayg | tamizh_geek: oh yeah! I remeber that one! | 14:42 |
clayg | wait... no that's a new one - but it does the same thing as the old one? | 14:43 |
*** peluse has quit IRC | 14:43 | |
acoles | clayg: lol | 14:43 |
clayg | it took a long time to load in my head - so it was acctually a pretty low rate | 14:44 |
acoles | clayg: it was genuine! | 14:44 |
clayg | kudos! | 14:44 |
acoles | i am discarding discard :) | 14:44 |
clayg | oh yeah that - we need a glossary! | 14:45 |
acoles | clayg: 'DiskFile for Dummies' available at any good bookshop | 14:45 |
*** delattec has quit IRC | 14:46 | |
*** cdelatte has quit IRC | 14:46 | |
clayg | acoles: that's funny | 14:47 |
*** mahatic has quit IRC | 14:47 | |
*** haomaiwang has quit IRC | 14:47 | |
*** mahatic has joined #openstack-swift | 14:47 | |
*** cdelatte has joined #openstack-swift | 14:48 | |
*** delattec has joined #openstack-swift | 14:48 | |
*** garthb has joined #openstack-swift | 14:50 | |
notmyname | acoles: does the table on https://wiki.openstack.org/wiki/Swift/PriorityReviews look reasonable to you? | 14:51 |
acoles | notmyname: mostly. i think https://review.openstack.org/#/c/213147 is going to be redundant once i am done with patch 215276 | 14:55 |
patchbot | acoles: https://review.openstack.org/#/c/215276/ | 14:55 |
acoles | notmyname: so i think we could avoid parallel/redundant effort there | 14:55 |
*** itlinux has joined #openstack-swift | 14:55 | |
notmyname | ok, great | 14:55 |
acoles | the optimistic gets will handle alt frags as a side effect of fetching durable frags when necessary | 14:56 |
notmyname | shall I drop patch 213147 references then? | 14:56 |
patchbot | notmyname: https://review.openstack.org/#/c/213147/ | 14:56 |
acoles | i should say.. "IMHO" | 14:56 |
notmyname | :-) | 14:56 |
clayg | notmyname: that's the same thing I said last night | 14:57 |
acoles | right now patch 215276 on gerrit doesn't do alt frags but it will | 14:57 |
patchbot | acoles: https://review.openstack.org/#/c/215276/ | 14:57 |
clayg | one patch with two closes-bug is better than two patches with one closes-bug each | 14:57 |
acoles | also i think clayg and peluse and i agree that alt frags is a less likely scenario | 14:57 |
acoles | clayg: i actually implemented frag tracking in the proxy today and discovered there are SO many tests that don't send a frag index with faked responses :/ | 14:58 |
acoles | so my poor proxy code amassed lots of frag index=None | 14:59 |
notmyname | priority reviews and dashboard updated | 14:59 |
notmyname | https://review.openstack.org/#/c/225357/ merged last night. does that affect the prioritization of https://review.openstack.org/#/c/211338 ? | 15:02 |
*** pberis has quit IRC | 15:04 | |
tamizh_geek | clayg: It fixes the same bug 1182155, but this will be a bigger change touching all over the code base. I'm just starting, but wanted to confirm if the direction I'm heading is the right one? | 15:05 |
openstack | bug 1182155 in OpenStack Object Storage (swift) "Error responses ignore accept header" [Undecided,In progress] https://launchpad.net/bugs/1182155 - Assigned to Azhagu Selvan SP (tamizhgeek) | 15:05 |
*** pberis has joined #openstack-swift | 15:05 | |
*** lpabon has joined #openstack-swift | 15:06 | |
*** minwoob has joined #openstack-swift | 15:06 | |
*** ahale has quit IRC | 15:07 | |
*** ahale has joined #openstack-swift | 15:07 | |
notmyname | clayg: I'm not getting any local unittest failures for https://review.openstack.org/#/c/222799 (I assume like you). but the gate certainly doesn't like it | 15:12 |
notmyname | tamizh_geek: I think there's a few things you can do to help out reviewers looking at your patches. first, make sure that the standard tests are passing. | 15:15 |
notmyname | tamizh_geek: if you get stuck on a particular test that's failing, that's a great thing to ask people for guidance or help on | 15:15 |
*** itlinux_ has joined #openstack-swift | 15:16 | |
notmyname | tamizh_geek: and then when you have the tests passing, a specific question like "I was considering A and B, I think A is better because..., what do you think?" is going to get a better review response than "did I do this whole thing right" | 15:16 |
notmyname | tamizh_geek: but in general, as we talked about previously, the general idea of what you're working on is good | 15:16 |
notmyname | tamizh_geek: thanks for tackling it | 15:17 |
tamizh_geek | notmyname: Yes sorry, I should have fixed them. Just thought of validating it with more than one core dev. Thanks for the suggestions. | 15:17 |
*** itlinux has quit IRC | 15:18 | |
notmyname | tamizh_geek: right now we're all working on the last set of priority patches that need to land for the next release (by the end of next week). So if you don't get a lot of attention on your patch yet, don't feel bad :-) | 15:18 |
notmyname | tamizh_geek: and if you want to help out with getting those last patches working, take a look at the stuff listed on https://wiki.openstack.org/wiki/Swift/PriorityReviews | 15:18 |
tamizh_geek | notmyname: Ah okay! | 15:18 |
tamizh_geek | will sure do :) | 15:20 |
*** garthb has quit IRC | 15:20 | |
*** garthb has joined #openstack-swift | 15:20 | |
*** kevinc___ has joined #openstack-swift | 15:25 | |
*** pberis has quit IRC | 15:29 | |
*** pberis has joined #openstack-swift | 15:31 | |
*** delattec has quit IRC | 15:31 | |
*** cdelatte has quit IRC | 15:31 | |
tamizh_geek | notmyname: https://review.openstack.org/#/c/177195 I see this in merge conflict and required for Liberty release(If I'm right), can I fix it and send patch to the same patchset? Or should I start a new one? | 15:32 |
notmyname | tamizh_geek: if you can resolve the merge conflict, then yes, send it to the same change-id (same patch, new patchset in gerrit) | 15:33 |
*** alejandrito has joined #openstack-swift | 15:33 | |
*** trifon has joined #openstack-swift | 15:37 | |
*** klrmn1 has joined #openstack-swift | 15:38 | |
*** klrmn1 has quit IRC | 15:39 | |
openstackgerrit | Tim Burke proposed openstack/swift: Add the ability to specify ranges for SLO segments https://review.openstack.org/211726 | 15:44 |
*** sayali has quit IRC | 15:55 | |
*** esker has joined #openstack-swift | 15:56 | |
notmyname | clayg: locally on my saio, if I use a seed of 1, I get the exact same results as CI sees with a seed of 2 on http://logs.openstack.org/99/222799/5/check/gate-swift-python27/dc1c354/console.html#_2015-09-23_08_06_35_111 | 15:57 |
openstackgerrit | Tim Burke proposed openstack/swift: Add the ability to specify ranges for SLO segments https://review.openstack.org/211726 | 16:00 |
*** garthb_ has joined #openstack-swift | 16:00 | |
*** garthb has quit IRC | 16:02 | |
*** itlinux_ has quit IRC | 16:06 | |
*** esker has quit IRC | 16:06 | |
*** sayali has joined #openstack-swift | 16:08 | |
*** resker has joined #openstack-swift | 16:12 | |
*** itlinux has joined #openstack-swift | 16:13 | |
*** rledisez has quit IRC | 16:14 | |
*** resker has quit IRC | 16:16 | |
*** chsc has joined #openstack-swift | 16:17 | |
*** peluse has joined #openstack-swift | 16:17 | |
*** amit213 has quit IRC | 16:17 | |
*** amit213 has joined #openstack-swift | 16:17 | |
openstackgerrit | John Dickinson proposed openstack/swift: Add the ability to specify ranges for SLO segments https://review.openstack.org/211726 | 16:24 |
*** klrmn1 has joined #openstack-swift | 16:24 | |
*** klrmn1 has quit IRC | 16:24 | |
notmyname | timburke: I rebased on top of master so the coverage thing is picked up and CI (or at least the 3rd party CI) passes ^ | 16:24 |
*** pberis has quit IRC | 16:28 | |
*** lpabon has quit IRC | 16:28 | |
*** janonymous has joined #openstack-swift | 16:28 | |
*** pberis has joined #openstack-swift | 16:29 | |
clayg | tamizh_geek: notmyname is correct - but I generally like the "push early" strategy for code review - it's always better to get some early validation before you're invested if you're unsure | 16:29 |
*** gyee has joined #openstack-swift | 16:29 | |
*** jordanP has quit IRC | 16:29 | |
*** garthb_ has quit IRC | 16:30 | |
clayg | tamizh_geek: that said - we barely keep up with all the code that *is* ready for review - much less stuff that's still WIP - so specific questions help alot - "I pushed up change xyz, it's not ready, what do you think about <link to hunk of diff> WRT doing a instead of b" | 16:30 |
*** garthb has joined #openstack-swift | 16:30 | |
clayg | i stared the change - but I won't get a chance to look at it until off hours | 16:30 |
notmyname | clayg: I'd like to help out on the ring validation, but I'm not really sure where to start with getting the CI unit test to pass | 16:32 |
clayg | notmyname: kota says it fails if you run it with tox? | 16:35 |
*** pberis has quit IRC | 16:35 | |
notmyname | yeah, I'm doing that now. passes with ./.unitttests | 16:35 |
clayg | notmyname: he also had a follow up change that with smaller seeds that he said was better? | 16:35 |
*** pberis has joined #openstack-swift | 16:36 | |
clayg | notmyname: yeah I mean those two tests that fail are storta stupid - I think the pre-existing one was trying to validate the printing of the balance warning and the exit code - but with a valid ring it wasn't unbalanced anymore | 16:36 |
notmyname | tox just passed for me | 16:36 |
clayg | notmyname: if there was a way to make a ring that would be for sure unbalanced we could probably just fix the test to do that and remove the other. | 16:37 |
notmyname | test_warn_at_risk and test_no_warn_when_balanced? I'm only seeing the failure in test_no_warn_when_balanced | 16:37 |
clayg | yeah - something like that - they were right next to each other and they do basically the same thing with a different seed | 16:39 |
clayg | it was probably a bad idea - but it WOMM and had nothing to do with what I was acctually trying to change. | 16:40 |
notmyname | ya | 16:40 |
clayg | ... it probably could use an explicit test for failing validation :\ | 16:40 |
clayg | notmyname: anyway - one of them pre-existed - and like I said it was really there I think to validate the error-printing and exit code | 16:41 |
peluse | ctennis: you around? | 16:41 |
notmyname | yeah | 16:41 |
clayg | ... if it's not obvious how to create a ring that will absolutely be un-balanced (at risk) we should probably mock it up | 16:41 |
ctennis | peluse: y | 16:42 |
*** pberis has quit IRC | 16:43 | |
*** sanchitmalhotra has joined #openstack-swift | 16:43 | |
*** pberis has joined #openstack-swift | 16:44 | |
notmyname | clayg: ok, I'll work on that | 16:45 |
*** pberis has quit IRC | 16:45 | |
*** pberis has joined #openstack-swift | 16:46 | |
*** delattec has joined #openstack-swift | 16:47 | |
*** cdelatte has joined #openstack-swift | 16:47 | |
openstackgerrit | John Dickinson proposed openstack/swift: Validate against duplicate part replica device assignment https://review.openstack.org/222799 | 16:49 |
*** delattec has quit IRC | 16:50 | |
*** cdelatte has quit IRC | 16:50 | |
notmyname | that has kota's patch squashed into it to unblock CI. still the ugly tests, but perhaps this can let us move forward (perfect is the enemy of good and all that) | 16:50 |
clayg | oh - does that work? | 16:51 |
notmyname | well the tests passed for jenkins in gerrit, so I think so | 16:51 |
notmyname | also passes locally | 16:51 |
*** aix has quit IRC | 16:53 | |
*** pberis has quit IRC | 16:55 | |
*** pberis has joined #openstack-swift | 16:56 | |
*** mfalatic has joined #openstack-swift | 17:02 | |
*** mfalatic has quit IRC | 17:03 | |
*** mfalatic has joined #openstack-swift | 17:04 | |
*** amoturi has joined #openstack-swift | 17:05 | |
*** esker has joined #openstack-swift | 17:15 | |
*** xnox has quit IRC | 17:17 | |
*** pberis has quit IRC | 17:20 | |
*** pberis has joined #openstack-swift | 17:21 | |
*** peluse has quit IRC | 17:21 | |
*** pberis has quit IRC | 17:23 | |
*** pberis has joined #openstack-swift | 17:23 | |
*** klrmn1 has joined #openstack-swift | 17:26 | |
*** garthb has quit IRC | 17:26 | |
*** garthb has joined #openstack-swift | 17:26 | |
*** peluse has joined #openstack-swift | 17:29 | |
peluse | x | 17:30 |
*** pberis has quit IRC | 17:35 | |
*** pberis has joined #openstack-swift | 17:36 | |
*** xnox has joined #openstack-swift | 17:36 | |
*** geaaru has quit IRC | 17:40 | |
*** pberis has quit IRC | 17:40 | |
*** pberis has joined #openstack-swift | 17:40 | |
*** bill_az has joined #openstack-swift | 17:41 | |
*** esker has quit IRC | 17:43 | |
notmyname | clayg: gate is unblocked for https://review.openstack.org/#/c/222799/ | 17:44 |
*** esker has joined #openstack-swift | 17:50 | |
*** pberis has quit IRC | 17:52 | |
openstackgerrit | Tim Burke proposed openstack/python-swiftclient: Make LengthWrappers resettable if their _readable is https://review.openstack.org/226897 | 17:53 |
*** pberis has joined #openstack-swift | 17:54 | |
*** Fin1te has joined #openstack-swift | 18:02 | |
*** thurloat_isgone is now known as thurloat | 18:02 | |
*** esker has quit IRC | 18:03 | |
*** garthb has quit IRC | 18:03 | |
*** itlinux_ has joined #openstack-swift | 18:03 | |
*** garthb has joined #openstack-swift | 18:03 | |
*** itlinux has quit IRC | 18:04 | |
*** annegentle has joined #openstack-swift | 18:13 | |
*** itlinux has joined #openstack-swift | 18:23 | |
*** delattec has joined #openstack-swift | 18:24 | |
*** cdelatte has joined #openstack-swift | 18:24 | |
*** itlinux_ has quit IRC | 18:25 | |
*** pberis has quit IRC | 18:27 | |
*** pberis has joined #openstack-swift | 18:28 | |
*** robefran has joined #openstack-swift | 18:32 | |
*** pberis has quit IRC | 18:32 | |
*** aix has joined #openstack-swift | 18:33 | |
*** pberis has joined #openstack-swift | 18:33 | |
*** pberis has quit IRC | 18:37 | |
*** pberis has joined #openstack-swift | 18:38 | |
robefran | Does anyone know how to access account_info from within object-server middleware ? Specifically I'm trying to find the bytes used in the account X-Account-Bytes-Used . I know this is a dumb idea, but I am not allowed to add middleware to the proxy-server pipeline in my deployment. I am however allowed to add middleware to a specific set of object-servers, but I can't figure out how to make an account HEAD request from the | 18:41 |
robefran | object-server pipeline.. | 18:42 |
*** amoturi has left #openstack-swift | 18:42 | |
*** esker has joined #openstack-swift | 18:44 | |
*** pberis has quit IRC | 18:45 | |
*** pberis has joined #openstack-swift | 18:46 | |
*** ccavanna has quit IRC | 18:46 | |
*** pberis has quit IRC | 18:48 | |
*** pberis has joined #openstack-swift | 18:49 | |
jrichli | robefran: you might take a look at swift/common/internal_client.py: An internal client that uses a swift proxy app to make requests to Swift. | 18:50 |
*** CaioBrentano has joined #openstack-swift | 18:50 | |
jrichli | robefran: and there is the method swift/proxy/controllers/base.py:def get_account_info(env, app, swift_source=None) | 18:50 |
jrichli | but that is on the proxy | 18:50 |
robefran | jrichli: what I'm struggling with is the app, when I am using the object-server pipeline instead of the proxy-server pipeline | 18:52 |
robefran | jrichli: for InternalClient, I've only found uses in object-expirer and object-reconciler. I haven't figured out how to invoke it from object-server middleware | 18:54 |
*** delattec has quit IRC | 18:54 | |
*** cdelatte has quit IRC | 18:54 | |
robefran | jrichli: trying to get account_info from the object-server is likely a big mistake, and I may need to convince my deployment to let me add proxy-server middleware, but I wanted to first see if it could be done in the object-server pipeline | 18:55 |
*** layzee has joined #openstack-swift | 18:55 | |
robefran | jrhichl: oops, that should be container-reconciler, not object-reconciler | 18:57 |
jrichli | robefran: this has exceeded my knowledge. I can poke around to see what I can find. | 18:59 |
*** pberis has quit IRC | 18:59 | |
robefran | jrichli: the object-expirer.conf and container-reconciler.conf refer to app:proxy-server, I may be able to fake up my own .conf for my middleware to refer to app:proxy-server, and then use InternalClient from within my middleware. | 19:00 |
*** pberis has joined #openstack-swift | 19:01 | |
hrou | robefran, yea we use the IC (internal client) from storlet middleware. In theory our code can reside on both the object and proxy (that is the middleware lives on both), but I'm not entirely sure it'll work (I'll check): https://github.com/openstack/storlets/blob/master/Engine/swift/storlet_gateway/storlet_docker_gateway.py#L469 The key is the proxy conf file's existence on the obj server. | 19:01 |
clayg | are we having fun yet? I can't tell :\ | 19:01 |
clayg | what's that middleware that spits out storage node urls from the ring? | 19:01 |
clayg | endpoints? | 19:02 |
clayg | someone should do a version of IC that can use that api instead of needed real rings on disk - basically a python "APIRing" class that supports all the iter_nodes stuff that IC needs | 19:02 |
robefran | hrou: do I need the entire proxy-server.conf, or can I just use a partial fake .conf for the InternalClient invoked by my middleware? | 19:02 |
*** cdelatte has joined #openstack-swift | 19:03 | |
*** delattec has joined #openstack-swift | 19:03 | |
hrou | clayg, robefran - interesting ! Yea I wonder if that's a good approach instead. | 19:03 |
clayg | hrou: yeah that'd be sweet | 19:04 |
clayg | hrou: it could get even better with like signed requests to the storage nodes and the ability to fill in stuff like error limiting data | 19:04 |
hrou | robefran, nah I think the key is you need the proxy.conf and the ring files to know where to look ; ) | 19:04 |
clayg | hrou: and a pony | 19:04 |
hrou | list_endpoints | 19:05 |
robefran | hrou: ahh.. the ring files, yes you need all the proxy stuff | 19:05 |
*** kairo has quit IRC | 19:08 | |
notmyname | if you're going to the tokyo summit and for some reason you haven't registered yet, do it now. prices are about to go up, and the discount codes are about to expire | 19:09 |
*** thurloat is now known as thurloat_isgone | 19:11 | |
*** janonymous has quit IRC | 19:19 | |
*** sanchitmalhotra has quit IRC | 19:20 | |
*** esker has quit IRC | 19:22 | |
*** cdelatte has quit IRC | 19:22 | |
*** delattec has quit IRC | 19:22 | |
robefran | hrou: object-expire on the object nodes uses InternalClient, so everything is there for InternalClient | 19:23 |
hrou | robefran, yea that's a really good point. | 19:24 |
robefran | hrou: I'm going to try IC from within my middleware. The only problem may be the time taken by the loadapp() within InternalClient __init__ | 19:25 |
*** delattec has joined #openstack-swift | 19:25 | |
*** cdelatte has joined #openstack-swift | 19:25 | |
*** pberis has quit IRC | 19:26 | |
*** pberis has joined #openstack-swift | 19:27 | |
*** pberis has quit IRC | 19:30 | |
*** pberis has joined #openstack-swift | 19:32 | |
*** annegentle has quit IRC | 19:32 | |
*** annegentle has joined #openstack-swift | 19:32 | |
*** pberis has quit IRC | 19:36 | |
*** pberis has joined #openstack-swift | 19:37 | |
*** trifon has quit IRC | 19:38 | |
*** pberis has quit IRC | 19:46 | |
*** pberis has joined #openstack-swift | 19:47 | |
*** pberis has quit IRC | 19:51 | |
*** pberis has joined #openstack-swift | 19:52 | |
*** trifon has joined #openstack-swift | 19:53 | |
*** pberis has quit IRC | 19:53 | |
*** pberis has joined #openstack-swift | 19:54 | |
*** delattec has quit IRC | 19:57 | |
*** cdelatte has quit IRC | 19:57 | |
*** delattec has joined #openstack-swift | 19:58 | |
*** cdelatte has joined #openstack-swift | 19:58 | |
*** esker has joined #openstack-swift | 19:59 | |
*** esker has quit IRC | 20:02 | |
*** pberis has quit IRC | 20:08 | |
*** pberis has joined #openstack-swift | 20:09 | |
openstackgerrit | Alistair Coles proposed openstack/swift: Enable object server to return non-durable data https://review.openstack.org/215276 | 20:10 |
acoles | peluse: ^^ this now does alt frags on same node, which means you are a co-author! | 20:10 |
clayg | acoles: nice work! | 20:10 |
acoles | clayg: and you, fab review thanks, but more great ideas than i could implement today | 20:11 |
clayg | acoles: no worries - you're a machine! | 20:11 |
*** pberis has quit IRC | 20:13 | |
*** pberis has joined #openstack-swift | 20:14 | |
acoles | notmyname: i'm just heading home, may be a few mins late for meeting | 20:17 |
*** kevinc___ has quit IRC | 20:17 | |
acoles | there is ameeting right? | 20:17 |
*** acoles is now known as acoles_ | 20:20 | |
*** acoles_ is now known as acoles | 20:20 | |
*** acoles is now known as acoles_ | 20:21 | |
*** delattec has quit IRC | 20:22 | |
*** cdelatte has quit IRC | 20:22 | |
*** Fin1te has quit IRC | 20:23 | |
notmyname | acoles_: yup. there's a meeting | 20:23 |
*** mahatic has quit IRC | 20:25 | |
*** pberis has quit IRC | 20:26 | |
*** cdelatte has joined #openstack-swift | 20:26 | |
*** delattec has joined #openstack-swift | 20:26 | |
*** robefran has quit IRC | 20:38 | |
*** Fin1te has joined #openstack-swift | 20:44 | |
clayg | when is the meeting? | 20:47 |
*** zhill has joined #openstack-swift | 20:48 | |
*** dustins has quit IRC | 20:50 | |
*** ho has joined #openstack-swift | 20:51 | |
notmyname | clayg: 9 minutes | 20:52 |
*** cutforth has joined #openstack-swift | 20:52 | |
*** kota_ has joined #openstack-swift | 20:53 | |
*** ChanServ sets mode: +v kota_ | 20:53 | |
notmyname | in the meeting we'll be going over the patches on https://wiki.openstack.org/wiki/Swift/PriorityReviews | 20:53 |
kota_ | good morning | 20:54 |
ho | good morning! | 20:54 |
notmyname | clayg: I just learned earlier today that (in the old gerrit view at least) if you hover your mouse over a vote, it will tell you who voted for it | 20:54 |
notmyname | kota_: ho: hello! | 20:54 |
ho | notmyname: kota_: hello and good morning | 20:55 |
notmyname | meeting time (1 minute) | 20:59 |
mattoliverau | morning | 21:00 |
*** agomez has joined #openstack-swift | 21:03 | |
*** acoles_ is now known as acoles | 21:03 | |
*** hrou has quit IRC | 21:04 | |
acoles | notmyname: in the old gerrit view you can set an option to have it show you the reviewer's name by the vote | 21:05 |
*** Fin1te has quit IRC | 21:12 | |
*** delattec has quit IRC | 21:15 | |
*** cdelatte has quit IRC | 21:15 | |
acoles | notmyname: settings->preferences->'Display Person Name In Review Category' | 21:17 |
*** cdelatte has joined #openstack-swift | 21:19 | |
*** esker has joined #openstack-swift | 21:20 | |
notmyname | yeah, I had that on for a while. made the display too noisy, I thought | 21:21 |
*** trifon has quit IRC | 21:23 | |
*** agomez has quit IRC | 21:25 | |
openstackgerrit | paul luse proposed openstack/swift: Better error handling for EC PUT path when client goes away https://review.openstack.org/211338 | 21:31 |
clayg | kota_: I think the tests in https://review.openstack.org/#/c/220059 might be easier to follow if you wrote them with specific object server responses - like in unit.proxy.controllers.test_obj.class TestECObjController(BaseObjectControllerMixin, unittest.TestCase): | 21:39 |
clayg | kota_: TestECObjController I ment | 21:39 |
*** minwoob_ has joined #openstack-swift | 21:42 | |
acoles | kota_: so in patch 215276 i need to add some checks that the responses we collect in a bucket are conistent e.g. same timestamp, same etag, same status code maybe...so would probably be a natural place to check the ec scheme | 21:42 |
patchbot | acoles: https://review.openstack.org/#/c/215276/ | 21:42 |
acoles | *consistent | 21:42 |
*** minwoob has quit IRC | 21:43 | |
acoles | kota_: the ECGetResponseBucket can get smarter | 21:43 |
kota_ | great comments, reading. | 21:46 |
*** minwoob_ is now known as minwoob | 21:46 | |
*** cdelatte has quit IRC | 21:53 | |
acoles | kota_: i think the 'hard' part with adding more to 215276 will be writing good tests. there are so many variables and i want to be sure we cover them all. ec scheme would be one more. | 21:54 |
*** esker has quit IRC | 21:55 | |
acoles | peluse: clayg just want to sanity check something with you... | 21:55 |
* mattoliverau is off to find breakfast | 21:55 | |
peluse | acoles: sure, I'm totally not sane if thats th e question | 21:55 |
notmyname | https://wiki.openstack.org/wiki/Swift/PriorityReviews is updated | 21:56 |
kota_ | acoles: agreed. writing good tests is sometimes hard. | 21:56 |
kota_ | so | 21:56 |
acoles | as a simplification (and boy did i need some) I am only making the obj server report alt frags on same node that are *durable* | 21:56 |
clayg | acoles: why is that simpler? | 21:57 |
acoles | i'm assuming that multiple frags on same node with no durable is very unlikely | 21:57 |
peluse | yeah, I had some similar decisions to make in the original one and opted for "hey, alt frags is sorta of a best effort thing so KISS rules for sure" | 21:57 |
clayg | acoles: if the gather_* methods return a list of all the frags the object server could "easily" just return x-alternative-timestamp-frags: ts#, ts#, ts# | 21:57 |
* notmyname goes afk for a bit | 21:58 | |
peluse | clayg: yeah, I thought you were looking to see that a while ago btu wondered why it never materialized, or maybe I dreamed about it | 21:58 |
acoles | clayg: you're right, tbh i got hung up on how to best ship all that info back to the proxy - like would a header with a json dict keyed with timestamp and having lists of frag indexes be OK? | 21:58 |
clayg | peluse: well it wasn't so clear in my head how it would work until I started to see acoles cahnge come to be | 21:58 |
*** esker has joined #openstack-swift | 21:59 | |
acoles | clayg: and then in proxy i think it was simpler if i could assume all alt frags were durable so the existence of one made a bucket 'viable' | 21:59 |
clayg | acoles: it'd be ok - but I think a simple list of serailized ts# would be ok - basicallyt he file names "%(timestamp)s#%(frag_index)" | 21:59 |
acoles | ^^ thats a bit hazy i admit | 21:59 |
clayg | acoles: I was sorta thinking when you poop out the gather'd sorted file list - before you go into "select the most recent bestish set of this crap that would make up a file" - you could just sorta pop out the ts#'s that the proxy says it already has streams setup for | 22:00 |
clayg | so then when you look at the databag the only thing you could possibly come up with is something that be useful? | 22:00 |
clayg | so step 1) sort out everything I have 2) remove what the proxy already has 3) return the most useful looking thing from what I left if any | 22:01 |
acoles | clayg: sendng filenames exposes diskfile implementation all the way to the proxy | 22:01 |
clayg | acoles: nah - it's just a serialization format for the idea of (timestamp, frag_index) | 22:01 |
* peluse has to join another meeting for a bit... | 22:02 | |
acoles | hmmm, ok, maybe | 22:02 |
clayg | if you like {<timestamp>: [<frag_index>, ...], ...} like *way* better than a csv that's ok I think | 22:02 |
*** bhakta has quit IRC | 22:02 | |
clayg | acoles: sorry to poop on things - but I really am thinking that the whole idea of durable just doesn't belong in the GET path | 22:03 |
clayg | like if you PUT a replicated object and only one copy survies to disk - you get a 500 - then if you GET and you happen to get that data - GREAT - if not - GREAT! | 22:03 |
peluse | oh man | 22:03 |
clayg | it's not crazy to imagine that if that one and only one copy of data that was saved and returned you a 500 - disk' DIES - that next time you GET you get something else | 22:03 |
clayg | it's like the 500'd PUT never happened | 22:04 |
*** esker has joined #openstack-swift | 22:04 | |
acoles | clayg: ? so if we get ndata frags and none have a durable we go ahead and return the data? | 22:04 |
clayg | even if you were able to successfully read it ONCE that ONE time | 22:04 |
kota_ | Ah, yeah, I got it about the ECGetResponseBucket | 22:04 |
*** proteusguy__ has quit IRC | 22:04 | |
clayg | so on a GET to EC data the proxy and object server can ignore the whole consistency engine problem and return the best thing they can make up - durable or not really doesn't matter that much - trust the consistency engine will make the object readable if it can | 22:05 |
*** proteusguy__ has joined #openstack-swift | 22:05 | |
*** bhakta has joined #openstack-swift | 22:05 | |
kota_ | hmm, quick look at 215276, seems easy way to make another validation. | 22:05 |
acoles | clayg: but if there is no durable then that object will eventualy be deleted, no? | 22:05 |
clayg | I mean you could look at if it's durable or not - but if you can rebuild it - it *should* be durable - we're working hard on the PUT path to make sure that we never return 200 on an object that we're not going to be able to reconstruct - so let's use that simplicity on GET | 22:05 |
MooingLemur | clayg: found it! I see why my db replicators kept coming back with 0 containers/accounts to sync | 22:06 |
clayg | acoles: but that's what i'm saying - same thing in the replciated case - if you only have one copy on disk you can serve it - not so crazy to think that disk may die and you loose it | 22:06 |
*** hrou has joined #openstack-swift | 22:06 | |
acoles | clayg: true | 22:06 |
MooingLemur | clayg: I was binding to a long-form IPv6 address instead of canonical | 22:06 |
clayg | do we really think that we'll get ndata frags down and not have a single durable out there? | 22:06 |
acoles | no | 22:06 |
acoles | not often | 22:06 |
clayg | acoles: well not often == sometimes at scale - but I acctually have a hunch that we may be closer to practically never if you consider only the cases where we responded 201 on the PUT | 22:07 |
clayg | MooingLemur: good catch! | 22:07 |
clayg | acoles: so... sorry | 22:08 |
clayg | acoles: you can ignore me | 22:08 |
clayg | acoles: but I think the list-of-things-that-is-not-helpful is a good strategy in general over make-specific-request-for-something-that-i-think-this-one-time-might-have-helped | 22:08 |
acoles | clayg: no. i wouldn't have asked if it wasn't something i'd been considering. | 22:08 |
MooingLemur | clayg: and whataremyips from swift.common.utils was using the literal bind_ip value rather than normalizing it. Somehow this wasn't a problem before 2.4.0. I could also solve this by having the container/account servers bind to :: rather than a specific IP. | 22:09 |
clayg | MooingLemur: amazing - so you're having some sucess running swift on all ipv6 deployment - that's sort of exciting! | 22:09 |
*** garthb has quit IRC | 22:10 | |
MooingLemur | clayg: yeah, everything swift-related is IPv6 enabled and all talking IPv6 except for its memcache client :/ | 22:10 |
*** garthb has joined #openstack-swift | 22:10 | |
clayg | MooingLemur: sweet! | 22:10 |
clayg | MooingLemur: what's memcache's problem :P | 22:10 |
openstackgerrit | Kota Tsuyuzaki proposed openstack/swift: Fix missing container update https://review.openstack.org/186735 | 22:11 |
MooingLemur | clayg: memcached supports listening on v6.. was doing that 7 years ago. memcache.py simply doesn't appear to support IPv6. | 22:11 |
clayg | torgomatic: so 3 tiers one tier has twice as much as the other two - 1, 1, 2 - what's the needed overload? I still don't intuit this very well :\ | 22:12 |
clayg | MooingLemur: oh - well that's totally fixable! | 22:12 |
torgomatic | clayg: I think it's 0.5 | 22:12 |
clayg | MooingLemur: you gunna fix it! | 22:12 |
torgomatic | er, wait... 1.0? | 22:12 |
torgomatic | yeah, I'll go with that | 22:13 |
acoles | clayg: so would proxy only send list-of-things-that-is-not-helpful to nodes that have something helpful i.e. something not on that list? | 22:13 |
MooingLemur | clayg: I mean swift.common.memcached. it even has AF_INET sprinkled around in it. I suppose I could give it a try. | 22:14 |
*** annegent_ has joined #openstack-swift | 22:14 | |
clayg | acoles: idk, i was thinking tactically - instead of stuffing the headers in the alt_node dict - maybe somehow the ResponseGatheringBrain could get a callback hook into the "make_backend_headers" | 22:14 |
clayg | and basically ever time you fire a request - you should down the current most up to date list-of-things-that-are-not-helpful | 22:15 |
*** esker has quit IRC | 22:15 | |
clayg | then it doesn't matter if it's a priarmy, or a handoff or an alt frag - as long as it's not responding with something I already have a stream for | 22:15 |
kota_ | mattoliverau: new patch set uploaded on 186735 | 22:16 |
*** openstackgerrit has quit IRC | 22:16 | |
acoles | clayg: yes. i'd also been thinking about making the alt node thing a callback so proxy does the right thing according to current state rather then state when it dumped out the alt nodes | 22:16 |
*** openstackgerrit has joined #openstack-swift | 22:16 | |
acoles | clayg: got it. | 22:16 |
clayg | :\ | 22:16 |
clayg | yeah - sorry - seriously - i feel like such an armchair qb even suggesting all this :\ | 22:17 |
acoles | no its cool | 22:17 |
clayg | s/armchair qb/heckler from peanut gallery/ if that makes more sense | 22:17 |
acoles | backseat driver? | 22:18 |
clayg | maybe there's a way to convey my guilt w/o euphemism even, not sure | 22:18 |
clayg | acoles: yes thank you! | 22:18 |
*** annegentle has quit IRC | 22:18 | |
*** esker has joined #openstack-swift | 22:19 | |
clayg | torgomatic: so but overload is amplified per device somehow like (1, 1), (1, 1), (2, 2) doesn't end up needing as much overload as (1,), (1,), (2,) somehow? | 22:19 |
torgomatic | it's how much extra each device is willing to take... to balance out a 1-1-2 situation, you need to boost it up to a 2-2-2 so nobody is more than 1/3 of the weight | 22:20 |
*** jlhinson has quit IRC | 22:20 | |
clayg | so but then (1, 1), (1, 1), (2, 2) *should* still need 1.0 to work :\ | 22:21 |
acoles | clayg: ok i'll go look at that. | 22:21 |
acoles | clayg: another thing i have in mind is not assuming that responses with same timestamp implies same etag. I think I need to make the buckets cope with multiple etags just to be safe. Or do we just let the decoder blow up? | 22:23 |
clayg | acoles: heh | 22:23 |
*** esker has quit IRC | 22:24 | |
clayg | acoles: I don't think anyone would notice if you grouped by timestamp instead of etag - we all seemed to think that was the easiest thing to reason about | 22:25 |
acoles | clayg: i am grouping by timestamp now. | 22:25 |
clayg | acoles: in some ways the decoder would be happiest grouped by etag - but we might all loose our minds - so maybe it's better to use timestamp and ignore consistency issue | 22:25 |
clayg | acoles: ok, try not think about the craziness then | 22:26 |
acoles | but i was thinking i can group by (timestamp, etag) | 22:26 |
acoles | at least for real responses, not for alt frags | 22:26 |
acoles | but then my mind drifts to (timestamp, etag, status code) | 22:27 |
acoles | and kota_ wants to add ec scheme :P | 22:27 |
kota_ | acoles: exactly :\ | 22:28 |
acoles | clayg: ok its late here, watch this space, your X-Don't-Be-Unhelpful list thing is growing on me. | 22:28 |
clayg | acoles: kota_: wfm - tuples can be pretty longish | 22:28 |
clayg | acoles: have a good night - thanks for being awesome | 22:29 |
acoles | clayg: can a header name have a ' ? :P | 22:29 |
acoles | as in Don't | 22:29 |
acoles | jk | 22:29 |
acoles | good night | 22:29 |
kota_ | acoles: good night! | 22:30 |
acoles | kota_: sorry, good morning for you! | 22:30 |
*** acoles is now known as acoles_ | 22:30 | |
*** wbhuber has quit IRC | 22:31 | |
*** jamielennox is now known as jamielennox|away | 22:35 | |
*** alejandrito has quit IRC | 22:36 | |
mattoliverau | Morning (I'm back) | 22:39 |
mattoliverau | acoles: night | 22:39 |
mattoliverau | kota_: cool will take a look | 22:40 |
*** annegent_ has quit IRC | 22:49 | |
*** jerrygb has quit IRC | 22:55 | |
*** km has joined #openstack-swift | 23:03 | |
*** annegentle has joined #openstack-swift | 23:04 | |
*** annegentle has quit IRC | 23:05 | |
*** kota_ has quit IRC | 23:06 | |
*** agomez has joined #openstack-swift | 23:12 | |
*** chsc has quit IRC | 23:12 | |
*** garthb has quit IRC | 23:13 | |
*** kei_yama has joined #openstack-swift | 23:24 | |
*** _hrou_ has joined #openstack-swift | 23:46 | |
*** minwoob has quit IRC | 23:48 | |
*** hrou has quit IRC | 23:49 | |
*** jamielennox|away is now known as jamielennox | 23:52 | |
*** peluse has quit IRC | 23:54 | |
*** peluse has joined #openstack-swift | 23:55 |
Generated by irclog2html.py 2.14.0 by Marius Gedminas - find it at mg.pov.lt!