| *** lpabon has quit IRC | 00:03 | |
| *** vinsh has quit IRC | 00:04 | |
| *** jamielennox is now known as jamielennox|away | 00:09 | |
| *** ozialien has joined #openstack-swift | 00:16 | |
| *** jrichli has joined #openstack-swift | 00:21 | |
| *** vinsh has joined #openstack-swift | 00:22 | |
| *** jamielennox|away is now known as jamielennox | 00:23 | |
| *** welldannit has joined #openstack-swift | 00:23 | |
| *** welldannit has quit IRC | 00:23 | |
| *** gyee has quit IRC | 00:30 | |
| *** ho has joined #openstack-swift | 00:32 | |
| *** zhill has quit IRC | 00:50 | |
| notmyname | EC update: https://gist.github.com/notmyname/ddde46b782957dec6675 | 00:51 |
|---|---|---|
| *** ChanServ changes topic to "Review Dashboard: http://goo.gl/uRzLBX | Overview Dashboard: http://goo.gl/2By1qv | Priority Reviews: https://wiki.openstack.org/wiki/Swift/PriorityReviews | EC status: https://gist.github.com/notmyname/ddde46b782957dec6675 | Logs: http://eavesdrop.openstack.org/irclogs/%23openstack-swift/" | 00:52 | |
| *** kota_ has joined #openstack-swift | 00:54 | |
| notmyname | kota_: ho: thanks for waking up early for the meeting today! At the summit, I do want to talk about possibilities of better times. or just better ways to incorporate everyone across the timezones | 00:56 |
| kota_ | notmyname: No worries, I often join the meeting. Because the EC status is important for my company, I have to watch it and I am very happy to talk the great member in the meeting, every time. | 01:03 |
| notmyname | kota_: great :-) | 01:03 |
| notmyname | kota_: I jsut pasted a status update in the channel topic message | 01:04 |
| kota_ | notmyname: And then, the disussion about possibilities of better time sounds good for my sleep :P | 01:04 |
| notmyname | heh | 01:04 |
| kota_ | notmyname: ok, thanks | 01:04 |
| notmyname | I've got to run | 01:07 |
| *** ozialien has quit IRC | 01:53 | |
| *** G________ has joined #openstack-swift | 01:53 | |
| *** kota_ has quit IRC | 02:01 | |
| *** zhill has joined #openstack-swift | 02:01 | |
| *** zhill has quit IRC | 02:06 | |
| *** ozialien has joined #openstack-swift | 02:14 | |
| openstackgerrit | Thiago da Silva proposed openstack/swift: Refactoring the PUT method https://review.openstack.org/156825 | 02:18 |
| *** reed has quit IRC | 02:20 | |
| tdasilva | notmyname: great summary of EC update, thanks for putting that together! | 02:27 |
| *** haomaiwang has joined #openstack-swift | 02:29 | |
| *** dmorita_ has quit IRC | 02:36 | |
| *** aswadr has joined #openstack-swift | 02:43 | |
| *** erlon has quit IRC | 02:51 | |
| *** yuan has joined #openstack-swift | 03:18 | |
| *** G________ has quit IRC | 03:29 | |
| *** Gue______ has joined #openstack-swift | 03:29 | |
| *** echevemaster has quit IRC | 03:34 | |
| *** greghaynes has quit IRC | 03:43 | |
| *** kota_ has joined #openstack-swift | 03:55 | |
| *** ozialien has quit IRC | 04:01 | |
| *** ozialien has joined #openstack-swift | 04:02 | |
| *** lpabon has joined #openstack-swift | 04:02 | |
| *** lpabon has quit IRC | 04:02 | |
| *** echevemaster has joined #openstack-swift | 04:03 | |
| *** jamielennox is now known as jamielennox|away | 04:06 | |
| *** ppai has joined #openstack-swift | 04:33 | |
| *** jrichli has quit IRC | 04:38 | |
| *** Gue______ has quit IRC | 04:52 | |
| *** Gue______ has joined #openstack-swift | 04:53 | |
| *** Gue______ has quit IRC | 05:04 | |
| *** bkopilov has quit IRC | 05:26 | |
| *** zaitcev has quit IRC | 05:32 | |
| *** ozialien has quit IRC | 05:32 | |
| *** greghaynes has joined #openstack-swift | 05:40 | |
| *** SkyRocknRoll has joined #openstack-swift | 05:44 | |
| *** SkyRocknRoll has joined #openstack-swift | 05:44 | |
| *** echevemaster has quit IRC | 06:16 | |
| *** srsakhamuri has quit IRC | 06:18 | |
| *** srsakhamuri has joined #openstack-swift | 06:19 | |
| *** mmcardle has joined #openstack-swift | 06:25 | |
| *** mmcardle has quit IRC | 06:30 | |
| *** pcaruana has quit IRC | 06:33 | |
| *** ppai has quit IRC | 07:03 | |
| *** ppai has joined #openstack-swift | 07:16 | |
| *** bkopilov has joined #openstack-swift | 07:31 | |
| *** haomaiwang has quit IRC | 07:32 | |
| *** Bsony has joined #openstack-swift | 07:45 | |
| *** chlong has quit IRC | 07:47 | |
| ho | notmyname: thanks for cosideration of the meeting time. I appreciate it :-) | 08:05 |
| *** jistr has joined #openstack-swift | 08:37 | |
| *** geaaru has joined #openstack-swift | 08:40 | |
| *** jordanP has joined #openstack-swift | 09:05 | |
| *** logan2 has quit IRC | 09:22 | |
| *** logan2 has joined #openstack-swift | 09:22 | |
| *** bkopilov has quit IRC | 09:23 | |
| openstackgerrit | Merged openstack/swift: Per-policy DiskFileRouter https://review.openstack.org/164380 | 09:31 |
| *** silor has joined #openstack-swift | 09:33 | |
| *** mmcardle has joined #openstack-swift | 09:47 | |
| *** Trozz has quit IRC | 09:50 | |
| *** acoles_away is now known as acoles | 09:51 | |
| *** Trozz has joined #openstack-swift | 09:53 | |
| *** Trozz is now known as Trozz_ | 09:55 | |
| *** Trozz_ is now known as Trozz | 09:55 | |
| *** nshaikh has joined #openstack-swift | 10:00 | |
| *** joeljwright has joined #openstack-swift | 10:08 | |
| *** silor has quit IRC | 10:28 | |
| *** nshaikh has quit IRC | 10:29 | |
| *** kota_ has quit IRC | 10:41 | |
| openstackgerrit | Alistair Coles proposed openstack/swift: Add swift-recon feature to track swift-drive-audit error count https://review.openstack.org/163889 | 10:44 |
| *** ppai has quit IRC | 10:58 | |
| *** pcaruana has joined #openstack-swift | 11:05 | |
| openstackgerrit | Alistair Coles proposed openstack/swift: Tighten up recon middleware unit tests https://review.openstack.org/165774 | 11:08 |
| openstackgerrit | Alistair Coles proposed openstack/swift: Add swift-recon feature to track swift-drive-audit error count https://review.openstack.org/163889 | 11:13 |
| *** stockpirate has joined #openstack-swift | 11:55 | |
| *** Slashme_ has joined #openstack-swift | 11:59 | |
| *** mmcardle has quit IRC | 12:00 | |
| Slashme_ | Hi, I'm trying to set size quota's on swift at a tenant level. I've used "swift post -m quota-bytes:<size>" and I can see the metadata but it is not enforced. Am I missing something ? | 12:02 |
| *** geaaru has quit IRC | 12:07 | |
| *** chlong has joined #openstack-swift | 12:13 | |
| *** geaaru has joined #openstack-swift | 12:20 | |
| *** chlong has quit IRC | 12:26 | |
| *** mahatic has joined #openstack-swift | 12:36 | |
| *** chlong has joined #openstack-swift | 12:42 | |
| *** silor has joined #openstack-swift | 13:03 | |
| *** rdaly2 has joined #openstack-swift | 13:08 | |
| *** jistr has quit IRC | 13:08 | |
| *** jistr has joined #openstack-swift | 13:09 | |
| *** mmcardle has joined #openstack-swift | 13:10 | |
| *** chlong has quit IRC | 13:21 | |
| cschwede | Slashme_: do you have the quota middleware enabled in your proxy server? | 13:23 |
| Slashme_ | container-quotas and account-quotas, yes | 13:26 |
| *** Gu_______ has joined #openstack-swift | 13:31 | |
| Slashme_ | Okay, nevermind, it just takes a long time to update the total size, so you can still upload during this time. | 13:32 |
| Slashme_ | Thanks for your help though | 13:32 |
| *** silor1 has joined #openstack-swift | 13:37 | |
| *** silor has quit IRC | 13:40 | |
| cschwede | Slashme_: np, you’re welcome. You’re right, there might be a delay the update of the currently used space and the reported space, which is due to the eventual consistency | 13:46 |
| openstackgerrit | Alistair Coles proposed openstack/swift: Enable in-process functional test policy to be configured https://review.openstack.org/159205 | 13:49 |
| *** lpabon has joined #openstack-swift | 13:56 | |
| *** silor has joined #openstack-swift | 13:58 | |
| *** Gu_______ has quit IRC | 13:58 | |
| *** silor1 has quit IRC | 14:01 | |
| *** vinsh has quit IRC | 14:10 | |
| *** erlon has joined #openstack-swift | 14:26 | |
| *** rdaly2 has quit IRC | 14:29 | |
| *** jrichli has joined #openstack-swift | 14:30 | |
| *** Guest___ has joined #openstack-swift | 14:34 | |
| *** Guest___ has quit IRC | 14:46 | |
| *** ChanServ sets mode: +v cschwede | 14:52 | |
| *** marcusvrn has joined #openstack-swift | 14:56 | |
| *** tsufiev_ has left #openstack-swift | 15:05 | |
| notmyname | good morning | 15:12 |
| peluse | mornin' | 15:16 |
| peluse | Per-policy DiskFileRouter landed on feaure/ec.... nice! | 15:18 |
| notmyname | cool! | 15:19 |
| acoles | morning. yes saw that landed, just rev'ing the per-policy classes patch (next one up chain) to simplify it | 15:20 |
| peluse | interesting.. | 15:21 |
| acoles | peluse: and adding back tests that got lost in the move somehow :? | 15:23 |
| *** vinsh has joined #openstack-swift | 15:25 | |
| peluse | acoles, hope its not something I did! | 15:27 |
| * peluse hasn't been on that chain yet though :) | 15:27 | |
| acoles | peluse: no all my fault just didn't copy the tests | 15:28 |
| peluse | key word: yes | 15:28 |
| peluse | err... yet :) | 15:28 |
| acoles | clayg: (when you're awake) i decided to get rid of ReplicationDiskFileManager - excess baggage, can always go back in in future | 15:29 |
| *** gyee has joined #openstack-swift | 15:37 | |
| *** reed has joined #openstack-swift | 15:37 | |
| *** SkyRocknRoll has quit IRC | 15:39 | |
| *** Bsony has quit IRC | 15:39 | |
| *** lpabon has quit IRC | 15:40 | |
| peluse | acoles, still there? | 15:45 |
| acoles | peluse: here | 15:46 |
| peluse | question that you might be able to answer on .meta and ssync.... | 15:46 |
| *** Trozz has quit IRC | 15:46 | |
| *** Trozz has joined #openstack-swift | 15:47 | |
| acoles | ok | 15:47 |
| peluse | without digging super deep, it looks like if I have a .data and a .meta that need to get ssync'd over to somewhere else, the .meta is put into headers and ends up in xattr on the .data on the other end?? | 15:47 |
| peluse | so .meta never gets actually copied over?? | 15:47 |
| acoles | in short, yes. | 15:48 |
| acoles | imho that is wrong - because the .data file on receiver ends up with timestamp of the .meta on sender | 15:48 |
| acoles | i have fixes for it in my fast-post work | 15:49 |
| peluse | ahhh OK, good to know I'm not seeing things. I hadn't looked before and just *assumed* .meta was copied over but couldn't find the code anywere :) | 15:49 |
| peluse | I see what you mean about the dates being wrong though | 15:50 |
| acoles | is that a problem for you? i.e. are you using fast-post? | 15:50 |
| peluse | nope, that makes life easier for me right now on the reconstructor | 15:50 |
| acoles | phew | 15:50 |
| peluse | :) | 15:51 |
| cschwede | acoles: does that mean we need 138498 merged asap as well? | 15:51 |
| cschwede | patch 138498 | 15:52 |
| patchbot | cschwede: https://review.openstack.org/#/c/138498/ | 15:52 |
| peluse | wow, hadn't seen that one | 15:52 |
| peluse | its independent of EC work (just FYI) | 15:53 |
| acoles | cschwede: 138498 is where i fixed it but that depends on the fast-post container update patch!! | 15:53 |
| cschwede | yeah, i know that, and that makes me wonder (if we need the fix for EC now) | 15:54 |
| acoles | i'd love to see it merged :D but i think we'll need to stay focussed on EC :( | 15:54 |
| peluse | we don't need it for EC, no | 15:54 |
| acoles | cschwede: i could separate out the ssync fix, but it only applies if using fast-post | 15:54 |
| cschwede | acoles: ok, now i got it. thx | 15:55 |
| *** geaaru has quit IRC | 15:59 | |
| *** welldannit has joined #openstack-swift | 16:01 | |
| *** Slashme_ has left #openstack-swift | 16:14 | |
| *** aswadr has quit IRC | 16:14 | |
| peluse | acoles, so given my revelation about .ts and .meta not actually being treated as files wrt ssync, I think the changes you made to fname_to_ts for EC were slightly off in that the function only ever needs to manipulate .data file names - for .ts and .meta it should just be a pass through | 16:15 |
| peluse | acoles, I've made the changes on the resconstructor patch and will point them out for you once I push it... no hurry to think about it this second just wanted to mention it since we talked about it yesterday | 16:16 |
| acoles | peluse: ok lets take a look once everything is stacked on one chain | 16:18 |
| peluse | sounds good | 16:19 |
| acoles | peluse: but, .ts files must be considered by ssync, in order to replicate deletes - on master, yield_hashes will yield a ts timestamp and the ssync missing_check finds that is newer than receiver, then ssync updates calls a send_delete | 16:20 |
| peluse | guess I should write some tests to make sure deletes and meta are properly sync'd via reconstructor.... | 16:20 |
| acoles | peluse: as for the meta's, need to think more but only matters with fast-post | 16:21 |
| peluse | acoles, so I was just looking at that and it appears its done via diskfile exception. Once sec, I'll point to the code I was looking at | 16:21 |
| acoles | yes, it is , but you only get to the place where it will try to open the file if the missing_check is performed using the .ts timestamp | 16:21 |
| acoles | (i think) | 16:21 |
| peluse | so on master, the send_list is a list of obj hashes and timestamps only | 16:22 |
| acoles | so if sender has t0.data and t2.ts and receiver has t1.data, then if sender yield_hashes ignores t2.ts and sends (t0, hash) the receiver will not detect anything missing?? | 16:23 |
| peluse | one sec... | 16:23 |
| acoles | so nothing for that hash gets into send_list | 16:23 |
| *** Guest__ has joined #openstack-swift | 16:25 | |
| peluse | wait, why would the sender yield_hashes ignore t2.ts? | 16:25 |
| acoles | peluse: well, thats what has always confused me about the change to yield_hashes, that t2.ts has no #frag_index to match the trailing_filter | 16:26 |
| * acoles is easily confused by that method | 16:26 | |
| peluse | so I'm just focusing on master now to try and understand how .ts is propogated | 16:26 |
| peluse | so forget Ec for a second :) | 16:27 |
| acoles | also, your statement above ".ts and .meta not actually being treated as files wrt ssync" made me think .ts would be ignored | 16:27 |
| acoles | notmyname: peluse just said 'forget EC' :D | 16:28 |
| peluse | heh | 16:28 |
| notmyname | tsk tsk | 16:28 |
| notmyname | ;-) | 16:28 |
| peluse | hey, can I call you real quick? | 16:28 |
| notmyname | me? sure | 16:28 |
| peluse | acoles actually | 16:28 |
| notmyname | oh, I see how it is | 16:28 |
| acoles | sure you have my number still | 16:28 |
| acoles | hey notmyname can join in ;) | 16:29 |
| notmyname | no, I'm ok | 16:29 |
| acoles | peluse: have my number? | 16:29 |
| * notmyname goes back to blasting music at high volumes | 16:29 | |
| peluse | maybe shoot it to me again, can't find it in my scrollback | 16:29 |
| *** dmorita has joined #openstack-swift | 16:32 | |
| *** Bsony has joined #openstack-swift | 16:32 | |
| *** tsg_ has joined #openstack-swift | 16:33 | |
| *** jrichli has quit IRC | 16:35 | |
| *** jrichli has joined #openstack-swift | 16:35 | |
| *** Guest__ has quit IRC | 16:39 | |
| *** sandywalsh has joined #openstack-swift | 16:40 | |
| *** thumpba has joined #openstack-swift | 16:46 | |
| *** Guest__ has joined #openstack-swift | 16:48 | |
| *** thumpba_ has joined #openstack-swift | 16:48 | |
| *** thumpba has quit IRC | 16:51 | |
| *** sandywalsh has quit IRC | 16:51 | |
| acoles | peluse: the extra ssync tests i wrote for fast post are here https://review.openstack.org/#/c/138498/9/test/unit/obj/test_ssync_sender.py | 16:54 |
| acoles | peluse: see class TestSenderAndReceiver | 16:55 |
| acoles | peluse: and guess what i have right there: "# TODO: add tests to exercise deleted diskfile cases" :) | 16:55 |
| acoles | peluse: i'll revisit them and see if they can be generalised and lifted over to EC or master | 16:56 |
| *** gyee has quit IRC | 16:56 | |
| *** kbee has joined #openstack-swift | 16:57 | |
| *** zhill has joined #openstack-swift | 16:59 | |
| *** zhill has quit IRC | 16:59 | |
| *** pcaruana has quit IRC | 17:05 | |
| *** jistr has quit IRC | 17:07 | |
| peluse | acoles, awesome thanks | 17:12 |
| *** sandywalsh has joined #openstack-swift | 17:12 | |
| *** mahatic_ has joined #openstack-swift | 17:17 | |
| *** Bsony has quit IRC | 17:17 | |
| *** Bsony has joined #openstack-swift | 17:19 | |
| *** mahatic has quit IRC | 17:20 | |
| notmyname | for those who didn't see it last night, I put an EC status update link in the channel topic | 17:22 |
| clayg | morning | 17:24 |
| clayg | acoles: peluse: I'm still working on the tests for the hash suffix support for multi-fi | 17:25 |
| acoles | clayg: ack | 17:25 |
| clayg | peluse: acoles: not sure if makes a difference, but I'm ending up finding that the per_fi part of hashes_per_fi really only applies to .data files | 17:26 |
| notmyname | tsg_: around? I have isa-l compiling questions | 17:27 |
| clayg | peluse: acoles: everything else about ssyncing suffixes looks like it's going to be the same - basically .ts .meta .durable are all expected to be equal accross all the fa's in the suffix | 17:27 |
| tsg_ | notmyname: sure | 17:27 |
| notmyname | tsg_: I'm sure I'm jsut missing some dev headers or something | 17:27 |
| notmyname | tsg_: https://gist.github.com/notmyname/e02a5b2249f68571fd3d | 17:28 |
| tsg_ | notmyname: what platform is this on? | 17:28 |
| tsg_ | notmyname: Mac? Linux? | 17:28 |
| notmyname | tsg_: Linux saio 3.13.0-46-generic #79~precise1-Ubuntu SMP Tue Mar 10 20:25:33 UTC 2015 x86_64 x86_64 x86_64 GNU/Linux | 17:28 |
| tsg_ | notmyname: OK | 17:29 |
| tsg_ | notmyname: just a moment .. checking in my precise VM | 17:29 |
| notmyname | ls | 17:29 |
| acoles | clayg: sounds reasonable | 17:30 |
| *** silor has quit IRC | 17:31 | |
| clayg | acoles: just saw your note about dropping the base class - sounds good | 17:32 |
| clayg | acoles: does that mean I need a rebase? Or have you not pushed it yet? | 17:32 |
| tsg_ | notmyname: can you check your yasm version - mine reads '1.2.0' | 17:32 |
| acoles | clayg: not pushed yet, just working on one thing - do you need it? i can push what i have if you do | 17:33 |
| notmyname | tsg_: ah. I didn't have it and installed the distro version. looks like I have 1.1.0-1 | 17:33 |
| clayg | acoles: nope - i'm good | 17:33 |
| acoles | clayg: ok will push by end of my day | 17:33 |
| notmyname | tsg_: where can I get a newer one, if that's the issue? | 17:33 |
| tsg_ | notmyname: that may be the problem but certainly will help to update .. checking that for you (don't recall having manually installed though) | 17:34 |
| tsg_ | s/may/may or may not* | 17:34 |
| *** zhill has joined #openstack-swift | 17:35 | |
| tsg_ | notmyname: I am using the Trusty version http://mirrors.kernel.org/ubuntu/pool/main/y/yasm/yasm_1.2.0-1ubuntu1_amd64.deb | 17:37 |
| tsg_ | notmyname: (no dependencies it being yasm :)) | 17:38 |
| notmyname | tsg_: worked. thanks :-) | 17:38 |
| tsg_ | awesome! :) | 17:38 |
| *** marcusvrn has quit IRC | 17:39 | |
| tsg_ | notmyname: we can request yasm 1.2.0 backport to precise if needed | 17:39 |
| notmyname | tsg_: ImportError: No module named pyeclib.ec_iface | 17:40 |
| tsg_ | notmyname: "sudo ldconfig" | 17:40 |
| tsg_ | notmyname: this is with pypi version of pyeclib correct? | 17:41 |
| notmyname | as opposed to? | 17:41 |
| notmyname | from source? | 17:41 |
| tsg_ | installing liberasurecode separately (from source, or debs) | 17:42 |
| tsg_ | and then pyeclib | 17:42 |
| tsg_ | shouldn't matter, just checking to make sure we have the right liberasurecode version .. | 17:42 |
| *** kbee has quit IRC | 17:43 | |
| notmyname | tsg_: my mistake. I had uninstalled pyeclib | 17:44 |
| tsg_ | notmyname: ah that should do it :) | 17:45 |
| tsg_ | notmyname: btw, swifterdarrell created .debs for all the latest pieces including python-pyeclib and liberasurecode1/-dev - you should be able to use those as well | 17:45 |
| notmyname | hmm..now where would I have installed jerasure? | 17:47 |
| notmyname | ie to validate that I don't have it on my system any more? | 17:47 |
| tsg_ | notmyname: "ldconfig -v | grep Jerasure" | 17:48 |
| tsg_ | (shouldn't appear in the list if you don't have it in the loader search path) | 17:48 |
| notmyname | it's there. where did it come from? pip? | 17:49 |
| tsg_ | notmyname: correct | 17:49 |
| notmyname | but uninstalling pyeclib doesn't remove it | 17:50 |
| tsg_ | notmyname: the pip version of pyeclib installs liberasurecode and jerasure2 as dependencies | 17:50 |
| notmyname | ah ok | 17:51 |
| tsg_ | notmyname: we'll get rid of that rev once we switch to using the trusty-backport for latest liberasurecode and jerasure2 that is available now | 17:51 |
| tsg_ | notmyname: I should say, we'll switch to pyeclib-only version on pypi once we have the rpms available (we already have .debs) | 17:52 |
| *** mmcardle has quit IRC | 17:54 | |
| notmyname | tsg_: awesome. all (local) validation passed. thanks a ton | 17:57 |
| tsg_ | notmyname: glad it is working! | 17:58 |
| tsg_ | notmyname: this is with 'isa_l_rs_vand' backend? | 17:58 |
| notmyname | yup | 17:58 |
| tsg_ | great! | 17:59 |
| *** joeljwright has quit IRC | 17:59 | |
| notmyname | and more importantly, I then added jerasure to the system, switched to jerasue_rs_vand and downloaded the file put there with isa-l | 17:59 |
| notmyname | result? | 17:59 |
| notmyname | md5sum /tmp/isa 100MB /tmp/jerasure | 17:59 |
| notmyname | 2f282b84e7e608d5852449ed940bfc51 /tmp/isa | 17:59 |
| notmyname | 2f282b84e7e608d5852449ed940bfc51 100MB | 17:59 |
| notmyname | 2f282b84e7e608d5852449ed940bfc51 /tmp/jerasure | 17:59 |
| *** reed has quit IRC | 17:59 | |
| notmyname | the 100MB is my local source file. the other 2 were download from saio | 17:59 |
| *** lpabon has joined #openstack-swift | 18:01 | |
| tsg_ | notmyname: we tried to make them compatible at liberasurecode level .. however this will work as long as 'w' (word size), 'hd' (hamming distance) values are at defaults (configurable at liberasurecode level) | 18:01 |
| notmyname | tsg_: or, more specifically, the same? | 18:02 |
| tsg_ | notmyname: correct :) | 18:02 |
| notmyname | or does that only work with the defaults? | 18:02 |
| notmyname | ok | 18:02 |
| tsg_ | notmyname: should be the "same" | 18:03 |
| tsg_ | notmyname: thankfully we are not exposing those details at the Swift level | 18:03 |
| notmyname | +1 | 18:03 |
| tsg_ | notmyname: in general though, the results may vary so not recommended to change the 'implementation' - I will put some time into documenting limitations in that area | 18:04 |
| tsg_ | notmyname: .. why we decided to embed the ec_type in the frag archive metadata | 18:05 |
| *** gyee has joined #openstack-swift | 18:06 | |
| *** lpabon has quit IRC | 18:07 | |
| clayg | torgomatic: when you're running your test in a bash for loop to make sure it's stable - are you totally doing it right - or doing it *way* wrong? | 18:21 |
| *** jordanP has quit IRC | 18:22 | |
| clayg | a'ight i'm heding into the office | 18:26 |
| *** Guest__ has quit IRC | 18:28 | |
| *** tab___ has joined #openstack-swift | 18:28 | |
| *** reed has joined #openstack-swift | 18:29 | |
| *** Nadeem has joined #openstack-swift | 18:29 | |
| *** jmacs has quit IRC | 18:49 | |
| peluse | clayg, thanks | 18:58 |
| *** Guest__ has joined #openstack-swift | 19:13 | |
| openstackgerrit | Alistair Coles proposed openstack/swift: Add fragment index support to obj server and ssync https://review.openstack.org/165188 | 19:13 |
| openstackgerrit | Alistair Coles proposed openstack/swift: Per-policy DiskFile classes https://review.openstack.org/165125 | 19:13 |
| openstackgerrit | Alistair Coles proposed openstack/swift: Diskfile decides if durable is written based on policy https://review.openstack.org/165208 | 19:13 |
| acoles | i hope i didn't that screw up | 19:14 |
| acoles | or even screw that up | 19:14 |
| acoles | clayg: you probably want to rebase your hash-suffix stuff on 165125, or maybe 165188 (next one up chain) | 19:15 |
| acoles | clayg: or just re-write the whole lot ! | 19:16 |
| *** Guest__ has quit IRC | 19:16 | |
| acoles | clayg: you there? | 19:23 |
| acoles | clayg: mattoliverau commented on patch 165125 that maybe we should go ahead and move DiskFileManager down the file, where you had left a TODO in the policy router patch. | 19:24 |
| patchbot | acoles: https://review.openstack.org/#/c/165125/ | 19:24 |
| acoles | clayg: what do you think? i'm wondering why you didn't do it before? | 19:25 |
| *** panbalag has joined #openstack-swift | 19:30 | |
| *** tab___ has quit IRC | 19:34 | |
| *** Guest__ has joined #openstack-swift | 19:39 | |
| acoles | jrichli: added some thoughts on policy-specific func tests on this trello card https://trello.com/c/lcto2x95/148-regular-functional-test-mods-to-support-different-policies. HTH | 19:44 |
| acoles | peluse: notmyname : ^^ fyi | 19:44 |
| clayg | it's very curious that common.utils.listdir - just stright up mutes ENOENT | 19:45 |
| clayg | and replaces it with an empty list | 19:45 |
| jrichli | acoles: thanks, i will take a look | 19:45 |
| clayg | acoles: probalby because i'm lazy? what are you doing still working? | 19:45 |
| clayg | acoles: just makes the diff bigger - attaching it lazily works just fine without a bunch of red and gree in the diff | 19:45 |
| acoles | clayg: idk it just gets dark without me noticing | 19:46 |
| clayg | acoles: after I get my tests passing I'll rebase - thanks for the updates | 19:46 |
| acoles | clayg: yup, thats what i wondered, so shall we leave it with less red/green for the moment? | 19:47 |
| clayg | acoles: my goal is to make the review against master for obj.diskfile look like "a few lines change on the replicated diskfile but are almost obviously correct; then a bunch of EC stuff I don't care about got added - +2" | 19:48 |
| acoles | clayg: exactly | 19:48 |
| clayg | then either at the back of the EC merge chain we have a bunch of moving common code into sane base class higharchy and tests don't have to move - or we do that after Kilo when the risk of regressing because of a refactor has more time to suss out missing tests w/o having to backport a bunch of fixes | 19:49 |
| acoles | rather than, whoa +1000/-900, move on to another review :) | 19:49 |
| clayg | acoles: ;) | 19:49 |
| * acoles is not speaking form personal experience | 19:50 | |
| acoles | clayg: wise words. | 19:51 |
| acoles | clayg: jrichli : ok i'm calling it a day | 19:53 |
| clayg | acoles: nice work brother! | 19:53 |
| clayg | acoles: peluse: I'm going to have the multi-hash-suffix change up and tested rebased against the end of the per-policy diskfile refactor before I quit for the day if it kills me ;) | 19:54 |
| clayg | acoles: peluse: sorry it's taken so long | 19:54 |
| clayg | tdasilva: as soon as I get that done I'll get back to helping you with the proxy stuff | 19:54 |
| jrichli | acoles: ok, have a good evening! | 19:55 |
| *** acoles is now known as acoles_away | 19:57 | |
| peluse | don't let it kill you :) | 20:03 |
| *** Guest__ has quit IRC | 20:05 | |
| openstackgerrit | Thiago da Silva proposed openstack/swift: Refactoring the PUT method https://review.openstack.org/156825 | 20:08 |
| tdasilva | cschwede: ^^^ tried to fix the issue you raised with copy handling...let me know what you think... | 20:09 |
| *** G________ has joined #openstack-swift | 20:21 | |
| *** mahatic_ has quit IRC | 20:35 | |
| notmyname | so now that the per-policy diskfilerouter patch has landed, what's the next one? wasn't there one on that chain yesterday? I don't see any dependencies listed any more | 20:55 |
| notmyname | dfg: redbo: glange: here's something that you guys might be interested in https://bugs.launchpad.net/swift/+bug/1434262 | 21:00 |
| openstack | Launchpad bug 1434262 in OpenStack Object Storage (swift) "Object server does not ignore a down container server on writes leading to higher overall cluster latency" [Undecided,New] | 21:00 |
| dmsimard | How do you define the maximum amount of partitions/partition % to be re-assigned with swift-ring-builder ? I've seen --reassign-max but that doesn't seem to work. | 21:00 |
| notmyname | dmsimard: one partition replica is reassigned per rebalance. but otherwise it's based on the balance dispersion and weights | 21:01 |
| *** patchbot has quit IRC | 21:01 | |
| dmsimard | Well say I'm doubling the size of the cluster and want to spread out the amount of partitions across multiple rebalances (without re-assigning 100% partitions) | 21:02 |
| *** G________ has quit IRC | 21:03 | |
| *** patchbot has joined #openstack-swift | 21:03 | |
| *** tsg_ has quit IRC | 21:03 | |
| *** panbalag has quit IRC | 21:03 | |
| notmyname | dmsimard: then you'll gradually increase the weight of the drives in the new capacity areas | 21:04 |
| dmsimard | notmyname: Right, I forgot about using the weight. Good idea. Thanks | 21:04 |
| peluse | notmyname, I think next up for EC will either be next in that chain (router was the start) or another that clayg is working on now - don't know if hes got a D on that one or not (refactoring some suffix hashes stuff) | 21:09 |
| notmyname | peluse: IIRC it was https://review.openstack.org/#/c/165125/ but the merged patch doesn't show any dependency now | 21:10 |
| mattoliverau | Morning | 21:10 |
| notmyname | hi mattoliverau | 21:10 |
| jrichli | morning | 21:16 |
| notmyname | torgomatic: did you have a chance to take a look at the range GET patch after we talked yesterday? the submanifest functest error | 21:21 |
| torgomatic | notmyname: still working out some of the other bugs in multirange | 21:22 |
| notmyname | ack | 21:24 |
| *** os1 has joined #openstack-swift | 21:27 | |
| os1 | Hi | 21:28 |
| os1 | Could someone verify their /var/log/messages file on a Swift server | 21:28 |
| os1 | And do a search for this text? | 21:28 |
| os1 | recon-cron: [Errno 13] Permission denied: '/var/lock/swift-recon-object-cron' | 21:29 |
| os1 | I think there is an issue with the way Swift is deployed by default. | 21:29 |
| peluse | notmyname, maybe because it merged? hell I dunno | 21:31 |
| notmyname | peluse: ya, I'm thinking it's a gerrit artifact | 21:36 |
| openstackgerrit | paul luse proposed openstack/swift: Erasure Code Reconstructor https://review.openstack.org/131872 | 21:41 |
| openstackgerrit | paul luse proposed openstack/swift: Multiple fragment Archive Index Support https://review.openstack.org/159637 | 21:41 |
| peluse | FYi ^^ I have to keep the MFI patch above rebased until its time for me to move over to the other patch chain. Big PITA but otherwise the ECrecon will get too far out of date and I'll die | 21:41 |
| os1 | Also, does anyone know whether 'swift' is supposed to be included in the 'lock' Unix group? | 21:48 |
| peluse | os1, are you doing a SAIO or regular cluster? Which set of directions are you following to install? | 21:49 |
| *** tsg has joined #openstack-swift | 21:49 | |
| openstackgerrit | paul luse proposed openstack/swift: wip: ec reconstructor probe test https://review.openstack.org/164291 | 21:49 |
| *** imkarrer has joined #openstack-swift | 21:57 | |
| *** jrichli has quit IRC | 21:58 | |
| *** tab_____ has joined #openstack-swift | 22:00 | |
| os1 | peluse : It's a regular cluster. | 22:01 |
| *** jamielennox|away is now known as jamielennox | 22:09 | |
| *** tab_____ has quit IRC | 22:32 | |
| *** Nadeem has quit IRC | 22:38 | |
| *** erlon has quit IRC | 22:41 | |
| *** os1 has quit IRC | 22:44 | |
| *** imkarrer has quit IRC | 22:52 | |
| *** welldannit has quit IRC | 23:08 | |
| *** kutija has quit IRC | 23:24 | |
| *** thumpba_ has quit IRC | 23:24 | |
| notmyname | torgomatic: ping (re single ranged GETs | 23:25 |
| torgomatic | notmyname: yes? | 23:25 |
| notmyname | torgomatic: good news. the failing test I saw yesterday isnt' related to SLOs | 23:25 |
| torgomatic | notmyname: do you know what is breaking it, then? | 23:26 |
| *** chlong has joined #openstack-swift | 23:27 | |
| notmyname | torgomatic: bad news. it's worse | 23:27 |
| notmyname | :-) | 23:27 |
| notmyname | so there's a TestFile.testRangedGets() method | 23:27 |
| notmyname | that's good | 23:27 |
| notmyname | and it passes | 23:27 |
| notmyname | except, the file size it tests is 10000 bytes. which is much less than one ec segment. those test pass for any value smaller than one ec segment size or for any even multiple of the ec segment size. and it hangs for every fractional multiple | 23:29 |
| notmyname | torgomatic: with my brute-force method eg | 23:29 |
| notmyname | def testRangedGets(self): | 23:29 |
| notmyname | #file_length = 10000 | 23:29 |
| notmyname | #file_length = 1048576 # ec segment size | 23:29 |
| notmyname | file_length = 1048576 * 1.5 | 23:29 |
| notmyname | file_length = int(file_length) | 23:29 |
| notmyname | torgomatic: so, no, I don't know *why* yet. but I know it doesn't have anything to do with SLOs | 23:29 |
| torgomatic | huh, I could have sworn I handled that | 23:30 |
| torgomatic | with unit tests to cover it, even | 23:30 |
| notmyname | I could swear that you didn't ;-) | 23:30 |
| notmyname | torgomatic: ya, you've got big numbers in https://review.openstack.org/#/c/163620/2/test/unit/proxy/controllers/test_base.py | 23:31 |
| torgomatic | eh, maybe my arithmetic is off by one | 23:32 |
| torgomatic | or by a constant factor | 23:32 |
| torgomatic | or possibly it's just what you get if you throw a childrens' math workbook into a blender | 23:32 |
| torgomatic | hard to say | 23:32 |
| *** shri has joined #openstack-swift | 23:32 | |
| shri | Hi, I am seeing LockTimeouts errors in the account-server. http://pastie.org/private/2ex88czdsyjuh7gzn5wlw | 23:34 |
| shri | Any ideas why this happens? | 23:34 |
| notmyname | torgomatic: I'm looking at your ECObjectController GET handler code block to see if there is something in there | 23:36 |
| notmyname | torgomatic: if I fix it, I'll push up a revision. but, fair warning, I'm on kid duty in about 30-45 minutes, so I might not finish | 23:36 |
| notmyname | what is the fragment_size? | 23:46 |
| notmyname | how does it relate to the segment size? | 23:51 |
| *** vinsh has quit IRC | 23:59 | |
Generated by irclog2html.py 2.14.0 by Marius Gedminas - find it at mg.pov.lt!