| *** catintheroof has quit IRC | 00:00 | |
| *** catintheroof has joined #openstack-swift | 00:00 | |
| *** catintheroof has quit IRC | 00:04 | |
| thurloat | I'm working on an upgrade path for a cluster to go from 1.13.1/2 to latest. From my reading, aside from some changes that will need to be made on the proxy server -- the only dangerous looking version is 2.2.2 with the data placement changes. | 00:05 |
|---|---|---|
| thurloat | I'm considering that being the only stopping point along the way for the object servers, going from 1.13 -> 2.2.2 -> 2.15. | 00:06 |
| notmyname | thurloat: nope. the short version, at the risk of being too flippant, is that there's nothing to worry about there and it will just get better | 00:06 |
| thurloat | well that sounds lovely. | 00:07 |
| notmyname | the biggest things to worry about will be config changes, including proxy pipeline changes | 00:08 |
| thurloat | yeah I have a bunch of proxy code to rewrite anyway, i'll make sure to cover that. | 00:08 |
| notmyname | eg see the notes in 2.4.0 | 00:08 |
| notmyname | dependency changes should all be called out for every release | 00:09 |
| notmyname | the post-as-copy vs fast-post change is something to pay attention to | 00:09 |
| notmyname | the stuff that was refactored to be put into middleware is something to pay attention to, although mostly it should "just work" | 00:10 |
| notmyname | chetna: what you're describing is correct | 00:11 |
| thurloat | in terms of deps, we'll not be doing an in-place SW update. we'll be deploying as a slug | 00:12 |
| notmyname | chetna: the swift api contract says that listings are returned in lexicographic order | 00:12 |
| notmyname | thurloat: I'm going to assume that's not "a snail without a shell" | 00:12 |
| notmyname | chetna: there's not a concept of an "invalid marker". a marker is simply a thing to sort before or after. so in your case, you got all 3 because they sort after aa and zz retunred a 204 (a special case for an empty result) | 00:13 |
| notmyname | chetna: for pagination, the recommended strategy is to do a listing, then do another with the last result from the previous listing as the marker. repeat until you get either zero results or less than "limit" results (defaults to 10k) | 00:14 |
| thurloat | notmyname: just my crummy nomenclature for a pre-compiled / assembled deployment. i think it came from working on google appengine many moons ago. | 00:17 |
| chetna | got you. Thanks @notmyname for explaining so well. | 00:17 |
| notmyname | chetna: please stay around and keep asking questions :-) | 00:18 |
| notmyname | chetna: how are you using swift? | 00:18 |
| thurloat | hopefully now that this cluster will be up to date, our changes and updates will be relevant and we can participate in upstream development again. | 00:19 |
| notmyname | thurloat: that's interesting. so swap out a machine at a time? sounds almost like a "take one down and replace with a completely new one" style of update | 00:19 |
| *** wasmum has quit IRC | 00:19 | |
| notmyname | thurloat: one key will be to *not* update config values to enable new functionality until after all the servers are upgraded | 00:20 |
| chetna | Thanks @notmyname. We am working on implementing swift like APIs in our product so have to check how Swift works and behaves. | 00:20 |
| notmyname | you don't want to have a new proxy take a request and an old proxy not able to serve it | 00:21 |
| notmyname | thurloat: an obvious example would be encryption | 00:21 |
| notmyname | thurloat: ..hmm.. fast-post may be something to pay attention to | 00:21 |
| notmyname | thurloat: note that the default for fast post has changed (but we haven't removed the old one yet!) | 00:22 |
| notmyname | thurloat: so, actually, there won't be a problem there. carry on | 00:22 |
| notmyname | (there is a slight change in behavior, but fast post enables *more* functionalty--a superset--so the upgrade won't break anything) | 00:22 |
| notmyname | chetna: that's ... interesting? you're making something to consume and use swift storage? or something that is a storage system that wants to pretend to be swift? | 00:23 |
| chetna | @notmyname: It is a storage system that pretends to be swift, facilitating our customers to use swift APIs to do CRUD operations. | 00:26 |
| notmyname | ah. what's it called? or what company is it from? | 00:26 |
| chetna | I work for ECS- Elastic Cloud Storage product of Dell EMC. | 00:27 |
| notmyname | ok | 00:27 |
| *** cshastri has joined #openstack-swift | 00:35 | |
| *** psachin has joined #openstack-swift | 00:38 | |
| openstackgerrit | Merged openstack/swift master: Clean up some leftover Python 2.6-isms. https://review.openstack.org/511988 | 00:42 |
| *** tovin07__ has joined #openstack-swift | 00:49 | |
| thurloat | notmyname: yeah we were going to hold off on encryption for now. | 00:51 |
| thurloat | thanks for all the insight. feeling pretty confident about the process | 00:52 |
| *** kiswe has joined #openstack-swift | 00:54 | |
| *** kiswe has left #openstack-swift | 00:54 | |
| thurloat | so if I'm reading that right, fast-post should be disabled / object_post_as_copy = True | 00:55 |
| *** gkadam has quit IRC | 00:57 | |
| notmyname | correct post as copy is the opposite of fast post | 00:58 |
| notmyname | you want fast post, eventually | 00:58 |
| *** gkadam has joined #openstack-swift | 00:58 | |
| thurloat | and once everything is equal, it should be good to toggle on | 01:06 |
| *** gkadam has quit IRC | 01:06 | |
| notmyname | yeah, that's definitely safest | 01:08 |
| thurloat | :thumbsup: | 01:09 |
| thurloat | think the proxy servers will get cranky at object servers being so different in versions? I assume rolling out object server updates should happen before proxy server updates. | 01:11 |
| *** klrmn has quit IRC | 01:14 | |
| *** catintheroof has joined #openstack-swift | 01:19 | |
| *** catintheroof has quit IRC | 01:21 | |
| tdasilva | kota_: hello, just FYI...I left a note for you here earlier today about maybe keeping history of swift3 files while copying the code over to the s3api branch, but while attempting the merging I was unable to keep the history | 01:34 |
| *** links has joined #openstack-swift | 01:44 | |
| *** two_tired has joined #openstack-swift | 01:45 | |
| *** links is now known as Jaison|away | 01:45 | |
| kota_ | tdasilva: thx, will take a look! | 01:46 |
| tdasilva | kota_: let me know how I can help | 01:46 |
| *** chlong has joined #openstack-swift | 01:49 | |
| *** chetna has quit IRC | 01:58 | |
| *** chetna has joined #openstack-swift | 01:58 | |
| openstackgerrit | Merged openstack/swift master: Respond 400 Bad Request when Accept headers fail to parse https://review.openstack.org/502845 | 02:06 |
| *** bkopilov_ has quit IRC | 02:12 | |
| *** bkopilov has quit IRC | 02:13 | |
| *** m_kazuhiro has joined #openstack-swift | 02:15 | |
| m_kazuhiro | good morning | 02:15 |
| *** klrmn has joined #openstack-swift | 02:24 | |
| * kota_ is back to online at his office | 02:30 | |
| kota_ | m_kazuhiro: morning | 02:30 |
| m_kazuhiro | kota_: morning! | 02:32 |
| *** tonyb has quit IRC | 02:36 | |
| *** tonyb has joined #openstack-swift | 02:37 | |
| * kota_ is scrolling back the irc logs... | 02:40 | |
| *** mat128 has quit IRC | 02:42 | |
| *** wes_dillingham has quit IRC | 02:43 | |
| kota_ | done to read | 02:46 |
| kota_ | tdasilva: at this point, I don't have strong opinion to keep the full swift3 git history on the swift repo. That is because it seems hard problem that we need infra team helps. | 02:48 |
| kota_ | tdasilva: and then, since we should move the directory from swift3 repo tree into swift repo (e.g. swift3/requests.py -> swift/common/middleware/s3api/requests.py) | 02:49 |
| kota_ | the history by line could be almost changed (e.g. the result of sort of 'git blame' shows just all of them a brand-new lines) | 02:50 |
| kota_ | to manage the older change before migrating work, the cutting release to 1.12 may be effective because we could work like as "if the change doesn't have history, look at the swift3 repo", maybe? | 02:56 |
| kota_ | and, I'm willing to keep soft-freeze at swift3 repo, after we did almost of the restructure work. | 02:57 |
| kota_ | tdasilva: (and any cores) Is it reasonable? | 02:57 |
| kota_ | k, I'll bring it the strategy on the next irc meeting. | 02:58 |
| *** Jaison|away is now known as links | 03:23 | |
| openstackgerrit | Matthew Oliver proposed openstack/swift master: Make max_ranges configurable https://review.openstack.org/493418 | 03:24 |
| *** gkadam has joined #openstack-swift | 03:33 | |
| *** bkopilov has joined #openstack-swift | 03:39 | |
| *** bkopilov_ has joined #openstack-swift | 03:40 | |
| *** mtreinish has quit IRC | 03:42 | |
| *** mtreinish has joined #openstack-swift | 03:42 | |
| *** tovin07__ has quit IRC | 03:45 | |
| *** tovin07__ has joined #openstack-swift | 03:46 | |
| *** tovin07 has joined #openstack-swift | 03:53 | |
| *** klrmn has quit IRC | 03:55 | |
| *** m_kazuhiro has quit IRC | 04:04 | |
| *** m_kazuhiro has joined #openstack-swift | 04:04 | |
| m_kazuhiro | clayg: Are you here? | 04:05 |
| kota_ | mattoliverau: btw, sorry I won't be able to attend tomorrow afternoon meeting for my fam reason. Instead, I'll make it in the thu morning meeting in our time. | 04:17 |
| mattoliverau | kota_: kk | 04:17 |
| mattoliverau | thanks for letting me know :) | 04:18 |
| kota_ | :) | 04:18 |
| m_kazuhiro | Umm clayg seems not in irc now. | 04:22 |
| m_kazuhiro | clayg: If you find this message, please check #6 on https://etherpad.openstack.org/p/swift_symlink_remaining_discussion_points . I heard from tdasilva that you requested POST query to symlink. I want to know what is use case for the query. | 04:25 |
| m_kazuhiro | clayg: So If you have concrete use cases. Please write them on the etherpad. | 04:31 |
| clayg | Only usecase I was aware of for putting metadata ON the symlink was internal. Like if some transparent to the client trick was using symlinks as an implementation detail. | 04:35 |
| clayg | i.e. versioning or tiering didn’t want to prevent a POST | 04:36 |
| clayg | I think the dangerous assumption with the 3xx redirect error or whatever we do will be consistent - if we allow a post to create meta at t3 on a data at t1 - what do we do after the fact when a symlink at t2 shows up after we’ve already accepted the POST | 04:38 |
| clayg | I think the most defendable behavior there is the .meta at t3 is void but I don’t really know. | 04:40 |
| m_kazuhiro | clayg: I'm reading your comments now... Please wait a little. | 04:42 |
| clayg | I can certainly add some of that to the ether pad tmrw - but I’m also happy with no api to add meta to symlink for now | 04:42 |
| clayg | timburke: dose versioning or data protection plans for symlink a need metadata ON the symlink our of the gate or POST to symlink can be done later? | 04:43 |
| m_kazuhiro | clayg: As you said, tiering don't want to prevent a POST. But tiering can resolve it without "symlink=post", because tiering middleware will be in after symlink in pipeline. And I think voiding .meta on symlink is difficult because of consistency with container DB's information. | 04:46 |
| clayg | Good point about containers - but if symlink is just middleware how do you prevent object server from applying POST? You have to let the POST happen - and the you *still* return the redirect? Seems cheap to add qs that does that but just doesn’t transform response status code. | 04:50 |
| clayg | Anyway I have no usecase for which to advocate symlink POST beyond the consistency engine must be correct and the api must be explainable. ;) | 04:51 |
| clayg | tdasilva: ^ cc no symlink post is ok by me as long as the rest is correct! | 04:52 |
| clayg | G’night | 04:52 |
| m_kazuhiro | clayg: In current symlink design, we DON'T prevent object server from applying POST. If client send POST to symlink, metadata update will be applyied to the symlink with 3xx response. | 04:53 |
| clayg | Wfm! | 04:53 |
| m_kazuhiro | clayg: Thank you for your reply. I think we need to get timburke's opinion. Good night! | 04:57 |
| *** m_kazuhiro_ has joined #openstack-swift | 05:02 | |
| *** chetna has quit IRC | 05:05 | |
| *** m_kazuhiro has quit IRC | 05:05 | |
| *** two_tired has quit IRC | 05:10 | |
| *** tovin07__ has quit IRC | 05:30 | |
| *** ChubYann has quit IRC | 05:36 | |
| *** dfflanders has joined #openstack-swift | 05:50 | |
| *** geaaru has quit IRC | 06:03 | |
| *** mabrams has joined #openstack-swift | 06:15 | |
| *** gabor_antal has quit IRC | 06:17 | |
| *** SkyRocknRoll has joined #openstack-swift | 06:27 | |
| *** kei_yama has quit IRC | 06:32 | |
| *** psachin has quit IRC | 06:38 | |
| *** pcaruana has joined #openstack-swift | 06:44 | |
| *** psachin has joined #openstack-swift | 06:51 | |
| *** hseipp has joined #openstack-swift | 06:51 | |
| *** SkyRocknRoll has quit IRC | 06:52 | |
| *** kei_yama has joined #openstack-swift | 06:57 | |
| *** gabor_antal has joined #openstack-swift | 06:57 | |
| *** bkopilov_ has quit IRC | 07:02 | |
| *** SkyRocknRoll has joined #openstack-swift | 07:05 | |
| *** tesseract has joined #openstack-swift | 07:16 | |
| *** m_kazuhiro_ has quit IRC | 08:04 | |
| acoles | good morning | 08:04 |
| *** cbartz has joined #openstack-swift | 08:05 | |
| *** geaaru has joined #openstack-swift | 08:20 | |
| *** oshritf has joined #openstack-swift | 08:29 | |
| kota_ | acoles: morning | 08:37 |
| acoles | kota_: hi | 08:38 |
| *** SkyRocknRoll has quit IRC | 08:39 | |
| *** SkyRocknRoll has joined #openstack-swift | 08:51 | |
| *** SkyRocknRoll has quit IRC | 09:04 | |
| *** cshastri has quit IRC | 09:39 | |
| openstackgerrit | HCLTech-SSW proposed openstack/swift master: Add ability to undelete an account. https://review.openstack.org/507808 | 09:42 |
| *** kei_yama has quit IRC | 09:44 | |
| *** NM has joined #openstack-swift | 09:51 | |
| *** mabrams has quit IRC | 10:05 | |
| *** mabrams has joined #openstack-swift | 10:06 | |
| *** mvk has quit IRC | 10:19 | |
| *** m_kazuhiro has joined #openstack-swift | 10:24 | |
| *** Jeffrey4l has quit IRC | 10:24 | |
| *** Jeffrey4l has joined #openstack-swift | 10:24 | |
| *** NM has quit IRC | 10:25 | |
| *** Jeffrey4l has quit IRC | 10:33 | |
| *** openstackgerrit has quit IRC | 10:33 | |
| *** Jeffrey4l has joined #openstack-swift | 10:33 | |
| *** tesseract has quit IRC | 10:43 | |
| *** tesseract has joined #openstack-swift | 10:43 | |
| *** bkopilov has quit IRC | 10:46 | |
| *** NM has joined #openstack-swift | 10:47 | |
| *** mvk has joined #openstack-swift | 10:50 | |
| *** NM has quit IRC | 10:52 | |
| *** oshritf has quit IRC | 11:02 | |
| *** oshritf has joined #openstack-swift | 11:04 | |
| *** m_kazuhiro has quit IRC | 11:10 | |
| *** mat128 has joined #openstack-swift | 11:55 | |
| *** chlong has quit IRC | 12:01 | |
| *** wes_dillingham has joined #openstack-swift | 12:11 | |
| *** bkopilov has joined #openstack-swift | 12:20 | |
| *** oshritf has quit IRC | 12:31 | |
| *** SkyRocknRoll has joined #openstack-swift | 12:33 | |
| *** gkadam has quit IRC | 12:36 | |
| *** gabor_antal_ has joined #openstack-swift | 12:37 | |
| *** gabor_antal has quit IRC | 12:38 | |
| *** SkyRocknRoll has quit IRC | 12:48 | |
| *** mjturek has joined #openstack-swift | 12:54 | |
| *** oshritf has joined #openstack-swift | 12:57 | |
| *** SkyRocknRoll has joined #openstack-swift | 13:01 | |
| *** SkyRocknRoll has quit IRC | 13:14 | |
| *** chlong has joined #openstack-swift | 13:42 | |
| mjturek | hey - I'm deploying with packstack and I noticed that the puppet-swift module produces a [filter:keystone] section in proxy-server.conf rather than [filter:keystoneauth]. Is [filter:keystone] a valid alias or something? | 13:49 |
| acoles | mjturek: as long as the filter name in the proxy server pipeline is also 'keystone' then it should be ok | 13:50 |
| *** psachin has quit IRC | 13:52 | |
| mjturek | acoles: just checked and it is, thanks for the help! | 13:52 |
| *** catintheroof has joined #openstack-swift | 13:53 | |
| *** Jeffrey4l has quit IRC | 13:56 | |
| *** Jeffrey4l has joined #openstack-swift | 13:59 | |
| *** catintheroof has quit IRC | 14:11 | |
| *** chetna has joined #openstack-swift | 14:18 | |
| *** chetna has quit IRC | 14:25 | |
| *** chetna has joined #openstack-swift | 14:25 | |
| *** chlong has quit IRC | 14:34 | |
| *** shaner has quit IRC | 14:37 | |
| *** Adri2000 has joined #openstack-swift | 14:39 | |
| *** links has quit IRC | 14:41 | |
| *** shaner has joined #openstack-swift | 14:47 | |
| *** chlong has joined #openstack-swift | 14:48 | |
| *** catintheroof has joined #openstack-swift | 14:51 | |
| *** NM has joined #openstack-swift | 14:53 | |
| *** klrmn has joined #openstack-swift | 14:57 | |
| *** catintheroof has quit IRC | 15:03 | |
| *** wes_dillingham has quit IRC | 15:17 | |
| *** mabrams has quit IRC | 15:25 | |
| *** psachin has joined #openstack-swift | 15:29 | |
| timburke | clayg: m_kazuhiro: my (admittedly, rather loose) plan for a symlink-backed versioned_writes would involve doing the POST and following any redirects in versioned_writes | 16:00 |
| *** pcaruana has quit IRC | 16:01 | |
| *** cbartz has quit IRC | 16:04 | |
| peluse | acoles, got your ears on? | 16:23 |
| acoles | peluse: hey | 16:24 |
| peluse | I see your name on the next IRC meeting chair, wondering if you'd entertain a short discussion on SPDK. Possible collab type stuff including an overview from the Huawei engineer that I mentioned last week. I don't know if she is available but I can check if you've got 10-15 min on the agenda... | 16:25 |
| *** hseipp has quit IRC | 16:25 | |
| peluse | Oct 18 or later... | 16:25 |
| *** psachin has quit IRC | 16:27 | |
| acoles | peluse: that'll be the *0700* UTC meeting we have every other week, next tomorrow, not the 2100 UTC meeting. Is 0700 UTC a good time for you? that's 2am I think for you? | 16:27 |
| peluse | ick :) | 16:27 |
| peluse | Maybe the next one after that then, LOL | 16:27 |
| acoles | peluse: AFAIK there will be a 2100 meeting tomorrow - looks like notmyname hasn't updated the agenda yet | 16:28 |
| *** vint_bra has joined #openstack-swift | 16:28 | |
| peluse | tomorrow is probably too soon, I don't know if she'll even see my email by then so whatever the next good (US time) date would be... | 16:28 |
| acoles | peluse: the 2100 meetings are weekly | 16:29 |
| peluse | OK great. Let me see when she is available and then I'll get back to you guys, thanks!! | 16:31 |
| acoles | peluse: awww, but 2am could be fun?? ;) | 16:32 |
| peluse | heh, that's ruin my week for sure! I'm shit without sleep :) | 16:32 |
| * peluse is waiting for some sort of smart ass remark from clayg as soon as he wakes up ;) | 16:33 | |
| notmyname | good morning | 16:33 |
| notmyname | oh hi peluse! | 16:33 |
| peluse | notmyname, morning! | 16:33 |
| peluse | trying to see if we can't get some bidness going between our communities! | 16:33 |
| acoles | notmyname: peluse was just asking if there were any reviews he could help out with :P ;-) | 16:33 |
| *** chsc has joined #openstack-swift | 16:33 | |
| *** chsc has joined #openstack-swift | 16:33 | |
| peluse | LOL! | 16:34 |
| peluse | +2 to everything | 16:34 |
| * acoles will always try to suck peluse back in | 16:34 | |
| notmyname | might be able to find one or two | 16:34 |
| peluse | oh wait, I can't do that anymore :( | 16:34 |
| acoles | peluse: it can be arranges | 16:34 |
| acoles | arranged* | 16:34 |
| peluse | lets meet up in Colorado to discuss... | 16:34 |
| notmyname | peluse: did you want to talk about spdk in the swift meeting tomorrow? | 16:35 |
| notmyname | IIRC there was a question on the openstack dev mailing list a few days ago about it | 16:35 |
| peluse | I can, sure. The dev who has a speicifc proposal might not be ready yet but I can at least give an overview and answer any initial questions on what it is/where it might fit | 16:36 |
| *** nikivi has quit IRC | 16:36 | |
| peluse | oh wait, that's 2:00 tomorrow my time I think right? I have a call I can't multi-task with but next week I can | 16:37 |
| notmyname | peluse: it's up to you. I don't know of anyone in our community talking about it. but if you our someone from your community wants to introduce something or ask questions or answer questions, that would be great. whenever you're ready | 16:37 |
| peluse | cool, I'll plan on coming Wed 25th then... | 16:38 |
| notmyname | ack. just remind me to add something to the agenda :-) | 16:38 |
| *** mvk has quit IRC | 16:38 | |
| peluse | notmyname, hey, add me to the agenda on the 25th will ya :) | 16:39 |
| notmyname | lol | 16:39 |
| *** vint_bra has quit IRC | 16:48 | |
| *** nikivi has joined #openstack-swift | 16:57 | |
| acoles | mattoliverau: kota_ mahatic rledisez cschwede there is nothing on the agenda for the 0700 meeting so let's skip it. I think I saw kota_ say he's not around too. | 16:59 |
| acoles | i.e. 0700 18th October tomorrow | 17:00 |
| *** tesseract has quit IRC | 17:02 | |
| *** caiobrentano has joined #openstack-swift | 17:03 | |
| *** openstackgerrit has joined #openstack-swift | 17:20 | |
| openstackgerrit | Alistair Coles proposed openstack/swift feature/deep: only make single request to get container listings https://review.openstack.org/512756 | 17:20 |
| *** mvk has joined #openstack-swift | 17:26 | |
| clayg | peluse: I can't do it with the *pressure* - what if my smart ass comment about my channel backlog blowing up and a hot air bag showing up is too edgy or sarcastic and I acctually offend someone! ;) | 17:27 |
| notmyname | torgomatic: I rebased patch 336323 locally. running tests. if they pass, I'll push it up (?) | 17:37 |
| patchbot | https://review.openstack.org/#/c/336323/ - swift - Add checksum to object extended attributes | 17:37 |
| *** geaaru has quit IRC | 17:47 | |
| *** mvk has quit IRC | 17:48 | |
| *** mvk has joined #openstack-swift | 17:49 | |
| *** shaner has quit IRC | 17:50 | |
| *** dfflanders has quit IRC | 17:50 | |
| *** gabor_antal_km has joined #openstack-swift | 17:53 | |
| *** gabor_antal_ has quit IRC | 17:53 | |
| openstackgerrit | John Dickinson proposed openstack/swift master: Add checksum to object extended attributes https://review.openstack.org/336323 | 18:06 |
| notmyname | torgomatic: WOMM ^ | 18:07 |
| openstackgerrit | Merged openstack/swift master: Allow SLOs to have zero-byte last segments. https://review.openstack.org/510716 | 18:08 |
| *** MVenesio has joined #openstack-swift | 18:11 | |
| MVenesio | Hi Guys i was wondering if in the new versions of swift is still a good strategy to deploy swift using one config file per each server disk for account container and objects and tunning rsync by port per disk for replication, or it doesn't make any sense now, and is the same to config all the disk under three single config files, uno for each account container and object services ? | 18:17 |
| *** shaner has joined #openstack-swift | 18:19 | |
| notmyname | I'd definitely want to run it by clayg, but IIRC, there haven't been any significant changes since those things were introduced what would suggest doing it any differently today | 18:20 |
| *** vint_bra has joined #openstack-swift | 18:24 | |
| MVenesio | notmyname: perfect so we can wait to clayg's input to be sure, thanks John | 18:25 |
| clayg | i love me some rsync_module per disk and servers_per_port settings | 18:26 |
| clayg | it's all about the concurrency per spindle - do everything you can to get there | 18:26 |
| *** shaner has quit IRC | 18:27 | |
| *** shaner has joined #openstack-swift | 18:28 | |
| clayg | swift is pretty flexible overall wrt how you do your configuration - conf.d or one config for separate replication network servers vs. cluster/proxy facing services... I can say what swiftstack does - but some of that is just because of "history" more so than "if I were deploying a new swift cluster today knowing everything I know about how stuff breaks in unexpected ways and ops has to jump in and get involved | 18:28 |
| clayg | in fires I would suggest swift be deployed exactly like XYZ | 18:28 |
| clayg | ... honestly I don't even think I could answer that definitively - but it'd be an interesting experience to discuss it with e.g. briancline and ahale and onovy and rledisez ... | 18:29 |
| notmyname | "What is the best way to deploy Swift, knowing everythign you know about when and where it breaks?" "Well, Swift is pretty good, but the *best* way? I've got a list of 10 names of people for you to hire." | 18:30 |
| *** ChubYann has joined #openstack-swift | 18:32 | |
| *** oshritf has quit IRC | 18:33 | |
| MVenesio | clayg: Yes the "per disk" configuration it's really tricky, with 20 or 30 configuration files per server, so for that reason if you say that don't really find a good reason to do it i'll prefer to have just 3 config files for all the disks, simplifying configuration and future administration. | 18:34 |
| *** oshritf has joined #openstack-swift | 18:35 | |
| notmyname | (actually, I'd have a list of up to 700 names) | 18:38 |
| MVenesio | notmyname: That's great, so far we've to still deploying by ourselves, but will be excellent to know about some new installation experiences, maybe briancline ahale or onovy can contribute with a few thoughts | 18:43 |
| notmyname | yeah, just a tongue-in-cheek way of saying that the best way is to be part of the community. which you've always been great at! :-) | 18:45 |
| MVenesio | totally agreed | 18:57 |
| MVenesio | thanks you guys @notmyname @clayg for your answers | 19:00 |
| *** NM has quit IRC | 19:03 | |
| *** NM has joined #openstack-swift | 19:07 | |
| openstackgerrit | Corey Bryant proposed openstack/swift master: Drop group comparison from drop_privileges test https://review.openstack.org/512797 | 19:08 |
| *** chetna has quit IRC | 19:10 | |
| openstackgerrit | Merged openstack/swift feature/deep: Include a shard's account in its name https://review.openstack.org/512380 | 19:13 |
| clayg | MVenesio: wait... why do you need 20-30 config files per server? I think we're missing something... | 19:21 |
| *** ukaynar has joined #openstack-swift | 19:22 | |
| openstackgerrit | Tim Burke proposed openstack/swift master: Add Preamble and Postamble to SLO and SegmentedIterable https://review.openstack.org/365371 | 19:23 |
| openstackgerrit | Tim Burke proposed openstack/swift master: Add base64decode function to common/utils https://review.openstack.org/512804 | 19:23 |
| clayg | I deploy with 2 config files per service (so [account,container,object]-server/[1,2].conf) then there's the rsync.conf.d - I *do* create a separate file new file for any disk I want to "close" with max_connections = -1 but I could do that with less files - it just simplified the management code to make it one file per disk... | 19:23 |
| MVenesio | clayg: if you have for example 24 disks to use, you have to create /etc/swift/object-server/1.conf to 24.conf | 19:24 |
| clayg | I think the author of patch 512797 is a new/drive-by contributor - but it sounds like getting that test cleanup merged sooner than later would be helpful | 19:25 |
| patchbot | https://review.openstack.org/#/c/512797/ - swift - Drop group comparison from drop_privileges test | 19:25 |
| clayg | MVenesio: that's... not true... at least not with newer swift versions - unless I don't understand what you're trying to accomplish? | 19:25 |
| MVenesio | clayg: that's the way we did some deployments with older versions one config file per disk with 1 worker set each and the same amount using r_1.conf to r_24.conf for replication | 19:27 |
| clayg | double check this https://docs.openstack.org/swift/latest/deployment_guide.html#running-object-servers-per-disk and then make sure you're using the interpolation values for rsync_module - the config option is setup to allow you to use a variable like literally `{replication_ip}` in the config value so the replication_ip of the node you're connecting to is interpolated into the configured template and used as | 19:28 |
| clayg | the rsync module name when connecting to the remote note | 19:28 |
| * clayg shrugs | 19:28 | |
| clayg | MVenesio: I mean that wfm - you can totally do that - KISS - love it! vive y deja vivir! | 19:30 |
| MVenesio | clayg: clayg: this is what i was talking about, i'll take a look at this doc thanks | 19:31 |
| *** silor has joined #openstack-swift | 19:32 | |
| *** silor has joined #openstack-swift | 19:33 | |
| *** NM has quit IRC | 19:42 | |
| *** NM has joined #openstack-swift | 19:45 | |
| *** catintheroof has joined #openstack-swift | 19:49 | |
| *** chetna has joined #openstack-swift | 19:57 | |
| *** ukaynar has quit IRC | 19:58 | |
| *** ukaynar has joined #openstack-swift | 19:58 | |
| *** silor has quit IRC | 20:00 | |
| *** ukaynar has quit IRC | 20:02 | |
| *** NM has quit IRC | 20:04 | |
| *** NM has joined #openstack-swift | 20:05 | |
| *** ukaynar has joined #openstack-swift | 20:06 | |
| openstackgerrit | Clay Gerrard proposed openstack/swift master: Drop group comparison from drop_privileges test https://review.openstack.org/512797 | 20:06 |
| *** chlong has quit IRC | 20:16 | |
| *** gupta18197 has joined #openstack-swift | 20:20 | |
| mattoliverau | \o/ it's a peluse! Hey man! | 20:22 |
| peluse | mattoliverau, what up man!? | 20:24 |
| *** catintheroof has quit IRC | 20:35 | |
| *** catintheroof has joined #openstack-swift | 20:36 | |
| *** catintheroof has quit IRC | 20:36 | |
| mattoliverau | Heaps :) glad to see you man, I hope things in your world are as awesome as you! Sorry it's still early here and running around dealing with a toddler and her breakfast. | 20:42 |
| peluse | mattoliverau, sounds like tons of fun.. yeah I'm gonna pop into next week's 21:00 IRC meeting and chat a little about what we're doing in the project I'm working on now and see if there some Swift type stuff we can do! | 20:48 |
| peluse | http://spdk.io | 20:48 |
| mattoliverau | Ahh cool! | 20:57 |
| *** MVenesio has quit IRC | 20:57 | |
| *** shaner has quit IRC | 21:05 | |
| *** mat128 has quit IRC | 21:07 | |
| *** NM has quit IRC | 21:09 | |
| *** caiobrentano has quit IRC | 21:16 | |
| *** caiobrentano has joined #openstack-swift | 21:17 | |
| *** chlong has joined #openstack-swift | 21:17 | |
| *** caiobrentano has quit IRC | 21:21 | |
| openstackgerrit | Tim Burke proposed openstack/swift feature/deep: Change ShardRange to take a created_at arg https://review.openstack.org/512867 | 21:23 |
| openstackgerrit | Tim Burke proposed openstack/swift master: Quiet test output when running test_utils.py in isolation https://review.openstack.org/512873 | 21:45 |
| *** shaner has joined #openstack-swift | 21:51 | |
| *** chetna has quit IRC | 21:52 | |
| *** chetna has joined #openstack-swift | 21:55 | |
| *** NM has joined #openstack-swift | 21:56 | |
| *** NM has quit IRC | 22:00 | |
| *** shaner has quit IRC | 22:08 | |
| *** shaner has joined #openstack-swift | 22:12 | |
| timburke | anyone else disquieted by that "code 400, message Bad HTTP/0.9 request type" message that shows up when you run ./.unittests ? | 22:14 |
| timburke | seems to be because of obj/test_server.py:TestObjectServer.test_multiphase_put_metadata_footer_disconnect | 22:14 |
| timburke | and it started happening with https://github.com/openstack/swift/commit/2a75091 -- but it's not at all clear to me *why* | 22:15 |
| openstackgerrit | Samuel Merritt proposed openstack/swift master: Fix some spelling in a docstring https://review.openstack.org/512884 | 22:16 |
| *** ukaynar has quit IRC | 22:21 | |
| clayg | torgomatic: ^ my bad! and thanks! | 22:22 |
| torgomatic | clayg: you're welcome | 22:22 |
| *** Adri2000 has quit IRC | 22:25 | |
| openstackgerrit | Tim Burke proposed openstack/swift master: Let clients request heartbeats during SLO PUTs https://review.openstack.org/509306 | 22:26 |
| *** wes_dillingham has joined #openstack-swift | 22:28 | |
| *** Harin has joined #openstack-swift | 22:32 | |
| *** Harin has quit IRC | 22:35 | |
| *** gupta18197 has quit IRC | 22:38 | |
| *** catintheroof has joined #openstack-swift | 22:43 | |
| drewn3ss | I have an issues in swift where I had a node reboot cause my swift disks to mount in different order. Is there a good way to clean that up? Will the replicators eventually drain the "wrong" directories and fill the intended directories under the new mountpoints? imagine /mnt1/objects/{0,1,2} and /mnt2/objects/{3,4,5} partitions became /mnt1/objects{0,1,2,3,4,5} and /mnt2/objects{0,1,2,3,4,5} and | 22:48 |
| drewn3ss | /mnt1/objects{0,1,2} are slowly filling from replication but /mnt1/objects{3,4,5} aren't really draining. | 22:48 |
| drewn3ss | is there a way to manually run the replication rsyncs, to move the data manually, or is there metadata somewhere that would go stale? | 22:50 |
| drewn3ss | secondarily, while I'm waiting for the above to work itself out via swift-object-replicator, can I increate partition-power on the rings, or would that be disasterous? | 22:51 |
| *** vint_bra has quit IRC | 22:53 | |
| drewn3ss | is there any way to check consistency of having a copy of all data accessible w/out brute-force pulling all of the data out of the API and validating all objects are intact? | 22:53 |
| *** vint_bra has joined #openstack-swift | 22:53 | |
| *** mjturek has quit IRC | 22:54 | |
| drewn3ss | I should mention, I'm on Mitaka | 22:55 |
| *** vint_bra has quit IRC | 22:58 | |
| notmyname | drewn3ss: that's not a swift-specific thing. just some "fun" that linux throws your way. you need to set a disk label and set up the fstab to mount by label or uuid | 23:08 |
| clayg | i wouldn't introduce a part power increase if you're already working through some replication wonkeyness (was part power increase even available in mitaka?) | 23:10 |
| *** catintheroof has quit IRC | 23:17 | |
| drewn3ss | notmyname: yeah, we changed to uuid mounting after finding this about two weeks too late. | 23:19 |
| drewn3ss | clayg: I think you can increase by 1 in mitaka, it seems. | 23:19 |
| drewn3ss | right now, my biggest problem is the partitions have 2.5mm files in them | 23:20 |
| drewn3ss | *facepalm* | 23:20 |
| drewn3ss | Which is quite related to https://bugs.launchpad.net/charm-swift-storage/+bug/1691570 | 23:20 |
| openstack | Launchpad bug 1691570 in OpenStack swift-storage charm "swift-object-replicator killing long-running rsync failing to actually kill rsync causing replication backlog" [Medium,Incomplete] | 23:20 |
| drewn3ss | yeah, I'm really afraid of the idea of doing any sort of rebalance operation on the rings with vast parts of partitions not living on the right disks | 23:24 |
| drewn3ss | though I guess there's no disk reassignment with part-power increase of 1 which is why I had to ask. | 23:24 |
| clayg | something like this might help you get an idea of how things are progressing... https://gist.github.com/clayg/5d3b3ca13dea15f6a48efaa3cea92d7d | 23:25 |
| clayg | drewn3ss: yeah but I think part-power increase effectively pauses rebalance until after you finish all the relinking... | 23:25 |
| *** kei_yama has joined #openstack-swift | 23:25 | |
| drewn3ss | ah, that would make a whole ton of sense. | 23:26 |
| drewn3ss | thanks for the script...will investigate and play with it a bit tomorrow. | 23:26 |
| clayg | GL | 23:31 |
| drewn3ss | clayg: thanks :) | 23:33 |
| *** chsc has quit IRC | 23:38 | |
| openstackgerrit | Merged openstack/swift master: Remove swift-temp-url script https://review.openstack.org/511981 | 23:57 |
| openstackgerrit | Merged openstack/swift master: Quiet test output when running test_utils.py in isolation https://review.openstack.org/512873 | 23:58 |
| openstackgerrit | Merged openstack/swift master: Fix some spelling in a docstring https://review.openstack.org/512884 | 23:58 |
Generated by irclog2html.py 2.15.3 by Marius Gedminas - find it at mg.pov.lt!