| *** saschpe has quit IRC | 00:00 | |
| *** saschpe has joined #openstack-swift | 00:06 | |
| *** matsuhashi has joined #openstack-swift | 00:29 | |
| clayg | torgomatic: well I don't know to solve the "can't delete until everything is enqueued" problem - but i'm pretty sure we should move him out of the way first thing so that he can't continue to poison the world, even if it means teaching another daemon to look for these moved-out-of-the-way-poisoned-dbs | 00:29 |
|---|---|---|
| clayg | idk, I'll try to reserve judgement until I finish looking at https://review.openstack.org/#/c/80731/ | 00:30 |
| clayg | nope, the change to the container server makes me feel like a move-out-of-way-for-async-drain would be a win on both sides, I think a fork out of the container-auditor would be a good plan | 00:37 |
| * clayg wonders why the container and obj updaters lazy load the ring... | 00:40 | |
| *** fifieldt has quit IRC | 00:47 | |
| portante | notmyname: you around? | 01:01 |
| portante | regarding atlanta summit | 01:01 |
| portante | are most folks flying in on monday, leaving friday? | 01:01 |
| clayg | portante: what time to keynotes start on Monday? I imagine some folks will show up on Sunday night | 01:02 |
| portante | yes, was wondering about that | 01:02 |
| clayg | it seems like the afternoon on the last day of the summit keeps getting pushed back later and later, i'd guess some folks will stick around until Saturday too :\ | 01:02 |
| clayg | we did that in HK - but that was partially because of the long flight home - the other part being... Friday night in HK! | 01:03 |
| portante | ;) | 01:03 |
| portante | so clayg, are you planning on attending atlanta? | 01:06 |
| portante | zaitcev: you around? | 01:06 |
| zaitcev | portante: yes | 01:06 |
| clayg | portante: yeah I'm planning on being there | 01:06 |
| *** d89 has joined #openstack-swift | 01:06 | |
| portante | are you flying in sunday, leaving sat? | 01:06 |
| portante | zaitcev, too? | 01:06 |
| zaitcev | portante: flying to where? | 01:07 |
| portante | atlanta | 01:07 |
| portante | or are you driving from across the country? | 01:07 |
| zaitcev | I don't have the budget and I decided not to pay my own way in this cycle | 01:07 |
| zaitcev | At least until we know if I can make hackathon | 01:07 |
| portante | clayg: did you already book your travel? | 01:13 |
| clayg | portante: not yet! | 01:14 |
| portante | dude, atc expires on the 21st | 01:14 |
| portante | ;) | 01:14 |
| clayg | portante: ? I got my atc tickets - just haven't booked flights/hotel? | 01:15 |
| portante | ah, cool | 01:15 |
| clayg | there's no rooms "at the conference" this year is there? it's like portland where everyone just gets rooms whereever and commutes/walks to the conference? | 01:16 |
| portante | the conf center has walkway to nearby hotels, they have rooms blocked at the westin and another place | 01:16 |
| clayg | portante: yeah it looks like we better get on that - but I don't think there's a deadline other than availibility... | 01:17 |
| clayg | oh no... i'm wrong there's a deadline | 01:17 |
| clayg | oh but it's in April not March | 01:17 |
| clayg | portante: YOU'RE FREAKING ME OUT MAN! | 01:17 |
| portante | so I booked the room for monday to friday, but yeah I am freaking out because fares start to rise soon | 01:18 |
| clayg | portante: you should book all the rooms and then scalp them | 01:18 |
| portante | I am cheap ol' hacker | 01:18 |
| portante | ;) | 01:18 |
| clayg | heh | 01:18 |
| *** fifieldt has joined #openstack-swift | 01:22 | |
| *** nosnos has joined #openstack-swift | 01:25 | |
| *** dvas_ has quit IRC | 01:31 | |
| openstackgerrit | Clay Gerrard proposed a change to openstack/swift: Fix changing of storage policy index https://review.openstack.org/72536 | 01:33 |
| clayg | torgomatic: how on earth did you ever figure out the ref syntax to push to gerrit outside of git review like that? did you have to like look at the git review source code? is it hiding in my .git/config somewhere? what does the "for/feature" part even mean? what's it look like if you wanted to do that for a change against master? | 01:35 |
| torgomatic | clayg: I worked with gerrit before it was cool ;) | 01:36 |
| clayg | whoa | 01:36 |
| clayg | aren't you on a train right now? | 01:36 |
| torgomatic | back in a previous job we used Gerrit, but we didn't have any fancy-pants "git-review" command, so we did it the hard way | 01:36 |
| torgomatic | clayg: welcome to THE FUTURE :) | 01:36 |
| torgomatic | but yeah, I've got about 5-10mins of tethering before my phone overheats and craps out | 01:36 |
| clayg | lol | 01:37 |
| torgomatic | so, to push for master, it's $ git push gerrit HEAD:refs/for/master/topicname (might also be refs/publish/master/topicname; I think one's deprecated) | 01:37 |
| clayg | i don't know how you find room to bust out your laptop on the train, I can't do it on the light-rail, but bart gets croweded too | 01:37 |
| torgomatic | clayg: the 6:30 train in the last car usually has seats; 6:15 or earlier is right out | 01:38 |
| *** chandan_kumar has joined #openstack-swift | 01:53 | |
| creiht | clayg: https://github.com/pandemicsyn/fgerrit/blob/master/README.md | 01:54 |
| * clayg prepares for a laugh | 01:55 | |
| creiht | it is actually quite awesome | 01:55 |
| *** shakamunyi has joined #openstack-swift | 02:00 | |
| *** nosnos_ has joined #openstack-swift | 02:16 | |
| *** nosnos has quit IRC | 02:19 | |
| *** gyee has quit IRC | 02:20 | |
| *** saschpe has quit IRC | 02:32 | |
| *** saschpe has joined #openstack-swift | 02:33 | |
| *** saurabh_ has joined #openstack-swift | 02:59 | |
| *** saurabh_ has joined #openstack-swift | 02:59 | |
| madhuri_ | clayg: Can you please have a look https://review.openstack.org/#/c/78530/ ? | 03:18 |
| madhuri_ | clayg: I want to provide fix for the incorrect behaviour. | 03:19 |
| *** chandan_kumar has quit IRC | 03:22 | |
| clayg | you mean where sending an empty header just ignores the header? object metadata is already "reset" on every "update" | 03:33 |
| clayg | I could imagine a client that is setting X-Meta-Magic: do-not-care and then checking "X-Meta-Magic" in resp.headers | 03:34 |
| clayg | in that case it would be a breaking api change if a similar client was using a newish curl with -H "X-Meta-Magic;" or a decent client that lets you send empty headers | 03:34 |
| clayg | madhuri_: be a decent thing to fix up in an api version bump though I think | 03:35 |
| clayg | I only mention it as an potentially breaking change because we had similar api cleanup's bite us in the past - swift's api has been stable for a LONG time anything that's client faceing and works today is probably being used by some client somewhere | 03:36 |
| clayg | you can put up the patch tho; if you were wondering if some core might reject it as change in the api tho... well that's probably warranted | 03:36 |
| clayg | dfg: glange: gholt: ^ re changing storing empty metadata headers on objects | 03:37 |
| madhuri_ | clayg: I was just woried about whether it would have any other bad impact. | 03:37 |
| clayg | i don't think anything in swift core relies on the behavior - hence the lack of testing for it ;) | 03:37 |
| clayg | but I was cool with the test being added since it would serve as a warning to anyone thinking about chaing it ;) | 03:38 |
| clayg | just didn't realize it'd be the same guy that wrote the test! | 03:38 |
| * clayg fully admits this is just one guys opinion | 03:38 | |
| clayg | best way to get consensus would be a working patch - not sure I can offer any guidance other than fair warning | 03:39 |
| clayg | it's probably a trivial change if you want to take a stab at it | 03:39 |
| madhuri_ | clayg: I am new to swift and was wondering whether such patch would be accepted or not? | 03:39 |
| clayg | any reason for doing so? I mean do you know of a client that's trying to do this and having a bad time because of the differences in container and object metadata? | 03:39 |
| madhuri_ | I am neither much aware whether it is incorrect behaviour or not. | 03:40 |
| clayg | madhuri_: I don't have a crystal ball (not trying to be snarky!) I just don't know how others would feel - my knee jerk having been bit in the past is "no backwards incompatible changes to client facing api behavior EVAR!!!" | 03:40 |
| madhuri_ | No, it's just that I wanted to contribute to swift ;) | 03:41 |
| clayg | madhuri_: if you're looking for something that folks would definately be receptive too I'd search on launchpad for open "bugs" with low-hanging-fruit tag - or just look for gaps in unit test coverage | 03:41 |
| madhuri_ | I would be really grateful to you for that :) | 03:42 |
| clayg | also you can review patches, every -1 "needs a better commit message" or -1 "this needs more tests" | 03:42 |
| clayg | ^ helps | 03:42 |
| madhuri_ | Yeah. I will try to do that too | 03:43 |
| madhuri_ | Thanks clayg in tonnes :) | 03:43 |
| clayg | this might work for the low-hanging-fruit tag -> http://goo.gl/x7ZD5S | 03:44 |
| clayg | benchmark it and have fun! | 03:44 |
| clayg | madhuri_: oh oh oh! you know one that I would just hug someone for? | 03:44 |
| clayg | we have a swift-obj-info tool in bin/ that's great for objects, and I wish we had a swift-db-info tool for containers/accounts | 03:45 |
| madhuri_ | clayg: Definiterly I will work upon it | 03:45 |
| clayg | esp when I'm trying to see what's going on with a replica that has a .pending file in place I always end up in a repl to commit_puts - what a pita - be great to have a helpful cli that lets me "for db in $(find /srv/node*/sd*/containers -name \*.db); do swift-db-info $db --commit; done" | 03:46 |
| madhuri_ | swift-obj-info what is it for? | 03:46 |
| clayg | inspecting metadata on a file on disk (mostly) | 03:46 |
| clayg | it has an option to do a md5 check | 03:46 |
| *** matsuhashi has quit IRC | 03:47 | |
| clayg | oh... i said "benchmark" that link - i ment "bookmark" - dunno if that was an obvious typo or not | 03:47 |
| madhuri_ | swift-obj-info, this idea sounds great to me. | 03:48 |
| clayg | hrm... there's acctually a lot less low-hanging-fruit that I realized - might need to do some bug triage :\ | 03:48 |
| madhuri_ | Will start working on it today itself. I mean now. | 03:48 |
| clayg | cool! | 03:48 |
| clayg | zaitcev: portante: does gluster have anything like swift-obj-info? Dunno if there's a better name for a container/account equivilant than swift-db-info? | 03:49 |
| madhuri_ | clayg: I must say you are a great advisor :) | 03:49 |
| clayg | i guess you could just make two tools swift-container-info and swift-account-info - probably be worth it to have most of the code rely on the broker anyways, then the bin script would just say: swift.common.utils.display_backend_info(AccountBroker|ContainerBroker) | 03:50 |
| clayg | gl! | 03:51 |
| zaitcev | clayg: sorry, I'm almost entirely clueless about what GlusterFS has in administration tool space. | 03:51 |
| zaitcev | Also, it's swift-object-info, not swift-obj-info. | 03:51 |
| * clayg hangs his head in shame | 03:52 | |
| madhuri_ | clayg: I have looked for it. But there is no such tool for Account/Container. | 03:52 |
| clayg | madhuri_: right! I was trying to trick you into writing one for me | 03:52 |
| madhuri_ | Sure. I would love to :) | 03:52 |
| zaitcev | yeah, it's sqlite3(1) time | 03:52 |
| clayg | zaitcev: well I was thinking he might get away just using the broker? | 03:53 |
| zaitcev | oh | 03:53 |
| clayg | most of the time my for loops just sqlite3 --line $db "select * from container_stat" | 03:53 |
| zaitcev | python and then import | 03:53 |
| clayg | ^ yeah that! | 03:53 |
| zaitcev | well, yes, that's what I do too | 03:53 |
| zaitcev | you have to find it first with swift-get-nodes | 03:54 |
| clayg | zaitcev: the part that gets annoying for me is when I want to also commit the .pending files; my oneliners start to take up many lines... | 03:54 |
| zaitcev | clayg: I see you're more serious about it than I ever was. | 03:55 |
| clayg | zaitcev: not serious enough to acctually write the tool - but apparently serious enough to try and hound poor madhuri_ into doing it! | 03:56 |
| madhuri_ | clayg: I will do it. It will be a good start for me. | 03:58 |
| madhuri_ | clayg: See you later. Going to have my breakfast. Thank you :) | 04:03 |
| *** saju_m has joined #openstack-swift | 04:12 | |
| *** haomaiwang has quit IRC | 04:20 | |
| *** haomaiwang has joined #openstack-swift | 04:20 | |
| *** otoolee has quit IRC | 04:22 | |
| *** otoolee has joined #openstack-swift | 04:22 | |
| clayg | hrmmmm... only managed to get 10 things into low-hanging-fruit - sorry for the lp email barrage | 04:25 |
| *** matsuhashi has joined #openstack-swift | 04:31 | |
| clayg | peluse: so that .pending_file change is turning out to be a real mess when it comes time to commit_puts - I'm going to try and benchmark a get_info call inside of put_object, but I imagine it'll be cheaper to hit the stats table during merge time (because the triggers are going to need it anyway) rather than during put_object (who currently doesn't even need to open the database) | 04:31 |
| clayg | benchmarking will tell... | 04:31 |
| *** haomaiw__ has joined #openstack-swift | 04:51 | |
| *** haomaiwang has quit IRC | 04:55 | |
| *** zaitcev has quit IRC | 04:56 | |
| madhuri_ | clayg: there? | 04:58 |
| clayg | i'ma say "no" | 04:59 |
| madhuri_ | Need some help | 05:00 |
| clayg | np, just ask | 05:00 |
| madhuri_ | swift-container-info must display all the information regarding a container. Right? | 05:00 |
| madhuri_ | Like no. of objects, ring location, bytes used etc.. | 05:01 |
| clayg | madhuri_: I'd just take everything that broker.get_info returns and put it on the screen | 05:01 |
| madhuri_ | This need to reed from container_stat table of container sqlite db file. | 05:01 |
| clayg | ^ side benifit of commint .pending | 05:01 |
| madhuri_ | Ok. | 05:01 |
| clayg | meh... try and use the broker for a first pass if you can | 05:02 |
| madhuri_ | .pending in sense, displaying all the pending entries in the file | 05:02 |
| clayg | oh... maybe - I was just thinking that .get_info is going to call commit_puts_stale_ok for you anyway - not sure there's an interface to turn that off or if the tool needs one | 05:03 |
| madhuri_ | And same for the account too? | 05:03 |
| clayg | i think that'd be ok | 05:03 |
| madhuri_ | But I can do that without using get | 05:03 |
| madhuri_ | *get_info method | 05:03 |
| madhuri_ | Rather I can use sqlite3 to do that for me | 05:04 |
| clayg | you might look at what get_info does - I think you'd end up writing more or less the same code that's already there - sorry if it seems trivial to just use the broker | 05:04 |
| madhuri_ | I have looked at it but I found the same problem _commit_puts_stale_ok() :( | 05:05 |
| clayg | what's the "probleme" - that it commits the .pending? that's like probably more than 50% of what I want - I'd consider a --no-commits option to be sorta "bonus" | 05:07 |
| clayg | but even then I might go after it by "broker = Broker(db_file); broker._commit_puts = lambda *args, **kwargs: None" | 05:08 |
| madhuri_ | So I will have to look for a way to turn it off when needed | 05:08 |
| clayg | monkey patch baby! | 05:08 |
| * clayg is shooting from the hip - you do whatever looks best to you | 05:09 | |
| madhuri_ | Ok. Got it | 05:09 |
| *** chandan_kumar has joined #openstack-swift | 05:10 | |
| *** gvernik has joined #openstack-swift | 05:10 | |
| madhuri_ | clayg: Thank you :) | 05:11 |
| clayg | i thought you were going to get something to eat? | 05:11 |
| *** gvernik has quit IRC | 05:12 | |
| madhuri_ | Ohh.. | 05:13 |
| madhuri_ | Wondering ever have you tasted Indian cuisines? | 05:13 |
| * clayg loves me some pakora | 05:14 | |
| clayg | or tandoori | 05:15 |
| *** chandan_kumar has quit IRC | 05:17 | |
| madhuri_ | clayg: Amazed :) | 05:19 |
| madhuri_ | Next time, I will surely bring you pakora ;) | 05:19 |
| *** nosnos has joined #openstack-swift | 05:19 | |
| *** nosnos_ has quit IRC | 05:23 | |
| clayg | pandemicsyn: I'm trying to benchmark this thing and got all my tcp connections ate up - so I changed /proc/sys/net/ipv4/tcp_tw_recycle to 1 <- is that insane? it sure let me keep opening connections! | 05:27 |
| clayg | redbo: ahale: ^ | 05:28 |
| *** Trixboxer has joined #openstack-swift | 05:29 | |
| clayg | peluse: oh god, put_object is like 1/4 of the speed with a get_info call stuck in there :\ - something like 70% slower :'( | 05:39 |
| clayg | fml | 05:39 |
| *** nosnos_ has joined #openstack-swift | 05:49 | |
| *** nosnos has quit IRC | 05:51 | |
| *** haomaiwa_ has joined #openstack-swift | 05:57 | |
| *** ppai has joined #openstack-swift | 05:57 | |
| *** haomaiw__ has quit IRC | 06:00 | |
| *** saju_m has quit IRC | 06:02 | |
| *** saju_m has joined #openstack-swift | 06:04 | |
| *** saju_m has quit IRC | 06:07 | |
| *** saju_m has joined #openstack-swift | 06:08 | |
| *** nshaikh has joined #openstack-swift | 06:10 | |
| *** d89 has quit IRC | 06:12 | |
| *** d89 has joined #openstack-swift | 06:16 | |
| clayg | sweet, I think i cracked it | 06:26 |
| clayg | notmyname: it's cool - storage policies is back on! | 06:26 |
| *** matsuhashi has quit IRC | 06:28 | |
| *** matsuhashi has joined #openstack-swift | 06:30 | |
| *** saju_m has quit IRC | 06:35 | |
| *** bvandenh has joined #openstack-swift | 06:59 | |
| *** bvandenh has quit IRC | 07:07 | |
| *** shakamunyi has quit IRC | 07:15 | |
| *** psharma has joined #openstack-swift | 07:18 | |
| openstackgerrit | Clay Gerrard proposed a change to openstack/swift: Container Confict on obj update with wrong Storage-Policy-Index https://review.openstack.org/79731 | 07:29 |
| *** bvandenh has joined #openstack-swift | 07:29 | |
| openstackgerrit | Clay Gerrard proposed a change to openstack/swift: Container Confict on obj update with wrong Storage-Policy-Index https://review.openstack.org/79731 | 07:34 |
| *** shakamunyi has joined #openstack-swift | 07:41 | |
| *** shakamunyi has quit IRC | 07:45 | |
| *** mrsnivvel has joined #openstack-swift | 07:51 | |
| *** saju_m has joined #openstack-swift | 07:56 | |
| *** tburnes has quit IRC | 08:08 | |
| *** tburnes has joined #openstack-swift | 08:09 | |
| *** Rock____ has joined #openstack-swift | 08:10 | |
| *** mmcardle has joined #openstack-swift | 08:26 | |
| *** nosnos has joined #openstack-swift | 08:33 | |
| *** nosnos_ has quit IRC | 08:37 | |
| *** shakamunyi has joined #openstack-swift | 08:42 | |
| *** shakamunyi has quit IRC | 08:46 | |
| *** mlipchuk has joined #openstack-swift | 08:49 | |
| *** nacim has joined #openstack-swift | 08:52 | |
| *** bada_ has quit IRC | 08:59 | |
| *** dvas has joined #openstack-swift | 08:59 | |
| *** bada_ has joined #openstack-swift | 09:00 | |
| *** dvas has quit IRC | 09:04 | |
| *** dvas has joined #openstack-swift | 09:15 | |
| *** tanee-away is now known as tanee | 09:21 | |
| *** tanee is now known as tanee-away | 09:21 | |
| *** tanee-away is now known as tanee | 09:21 | |
| *** jamieh has joined #openstack-swift | 09:36 | |
| *** jamieh is now known as Guest94414 | 09:36 | |
| openstackgerrit | A change was merged to openstack/swift: Merge branch 'master' into feature/ec https://review.openstack.org/81345 | 09:38 |
| openstackgerrit | A change was merged to openstack/swift: Store policy index in container_stat table https://review.openstack.org/71704 | 09:38 |
| openstackgerrit | A change was merged to openstack/swift: move cors-test-page to literal include https://review.openstack.org/81168 | 09:38 |
| *** dvas has quit IRC | 09:42 | |
| *** shakamunyi has joined #openstack-swift | 09:43 | |
| *** dvas has joined #openstack-swift | 09:44 | |
| *** d89 is now known as Dharmit | 09:47 | |
| *** shakamunyi has quit IRC | 09:47 | |
| *** Guest56062 has quit IRC | 09:51 | |
| *** taras_ has joined #openstack-swift | 09:51 | |
| *** dvas has quit IRC | 09:56 | |
| *** dvas has joined #openstack-swift | 09:56 | |
| *** nosnos has quit IRC | 09:59 | |
| openstackgerrit | Clay Gerrard proposed a change to openstack/swift: Container Confict on obj update with wrong Storage-Policy-Index https://review.openstack.org/79731 | 10:02 |
| *** nosnos has joined #openstack-swift | 10:09 | |
| *** bada_ has quit IRC | 10:15 | |
| *** Rock____ has quit IRC | 10:19 | |
| portante | clayg: yes, I think it does | 10:31 |
| *** shakamunyi has joined #openstack-swift | 10:43 | |
| *** shakamunyi has quit IRC | 10:48 | |
| *** fifieldt has quit IRC | 10:49 | |
| openstackgerrit | Clay Gerrard proposed a change to openstack/swift: Add "If-None-Match: *" support to PUT https://review.openstack.org/81646 | 10:52 |
| *** chandankumar_ has quit IRC | 11:00 | |
| *** chandankumar has joined #openstack-swift | 11:00 | |
| *** dvas has quit IRC | 11:01 | |
| *** matsuhashi has quit IRC | 11:01 | |
| *** dvas_ has joined #openstack-swift | 11:02 | |
| *** matsuhashi has joined #openstack-swift | 11:24 | |
| portante | clayg: go to sleep man! | 11:27 |
| *** a1|away is now known as JelleB | 11:29 | |
| *** dvas_ is now known as dvas | 11:36 | |
| *** shakamunyi has joined #openstack-swift | 11:44 | |
| *** shakamunyi has quit IRC | 11:48 | |
| *** fifieldt has joined #openstack-swift | 11:51 | |
| *** matsuhashi has quit IRC | 12:08 | |
| *** saju_m has quit IRC | 12:09 | |
| *** fifieldt has quit IRC | 12:10 | |
| *** ppai has quit IRC | 12:10 | |
| *** matsuhashi has joined #openstack-swift | 12:12 | |
| *** dvas has quit IRC | 12:12 | |
| *** matsuhashi has quit IRC | 12:18 | |
| *** JuanManuelOlle has joined #openstack-swift | 12:19 | |
| *** Dharmit has quit IRC | 12:23 | |
| *** ppai has joined #openstack-swift | 12:24 | |
| *** matsuhashi has joined #openstack-swift | 12:24 | |
| *** saju_m has joined #openstack-swift | 12:26 | |
| *** jairo has left #openstack-swift | 12:40 | |
| *** shakamunyi has joined #openstack-swift | 12:45 | |
| *** nosnos has quit IRC | 12:46 | |
| *** tongli has joined #openstack-swift | 12:47 | |
| *** matsuhashi has quit IRC | 12:47 | |
| *** shakamunyi has quit IRC | 12:49 | |
| *** saju_m has quit IRC | 12:53 | |
| *** piyush1 has joined #openstack-swift | 13:05 | |
| *** saju_m has joined #openstack-swift | 13:06 | |
| *** tdasilva has joined #openstack-swift | 13:09 | |
| *** zul has quit IRC | 13:12 | |
| *** zul has joined #openstack-swift | 13:15 | |
| *** psharma has quit IRC | 13:16 | |
| *** changbl has quit IRC | 13:16 | |
| *** mmcardle has quit IRC | 13:16 | |
| *** ppai has quit IRC | 13:21 | |
| *** saju_m has quit IRC | 13:23 | |
| *** shakamunyi has joined #openstack-swift | 13:32 | |
| *** dmsimard has joined #openstack-swift | 13:33 | |
| *** ppai has joined #openstack-swift | 13:33 | |
| *** PradeepChandani_ has joined #openstack-swift | 13:33 | |
| *** mkollaro has joined #openstack-swift | 13:36 | |
| *** PradeepChandani has quit IRC | 13:36 | |
| *** saju_m has joined #openstack-swift | 13:39 | |
| *** mmcardle has joined #openstack-swift | 13:40 | |
| *** PradeepChandani_ has quit IRC | 13:44 | |
| *** piyush1 has quit IRC | 13:45 | |
| *** PradeepChandani_ has joined #openstack-swift | 13:45 | |
| *** byeager has joined #openstack-swift | 13:46 | |
| *** byeager has quit IRC | 13:46 | |
| *** byeager_ has joined #openstack-swift | 13:46 | |
| *** saju_m has quit IRC | 13:48 | |
| *** jergerber has joined #openstack-swift | 13:58 | |
| luisbg | morning | 14:06 |
| *** piyush1 has joined #openstack-swift | 14:14 | |
| *** tanee is now known as tanee-away | 14:20 | |
| *** ppai has quit IRC | 14:22 | |
| *** tanee-away is now known as tanee | 14:25 | |
| *** judd7_ has joined #openstack-swift | 14:37 | |
| *** judd7_ has quit IRC | 14:42 | |
| *** judd7_ has joined #openstack-swift | 14:44 | |
| *** dmsimard1 has joined #openstack-swift | 14:58 | |
| *** tanee is now known as tanee-away | 14:58 | |
| *** dmsimard has quit IRC | 14:59 | |
| *** judd7_ has quit IRC | 15:03 | |
| openstackgerrit | Babu Shanmugam proposed a change to openstack/swift: RADOS (CEPH) DiskFile Backend https://review.openstack.org/60215 | 15:07 |
| *** judd7_ has joined #openstack-swift | 15:12 | |
| *** changbl has joined #openstack-swift | 15:18 | |
| *** d89 has joined #openstack-swift | 15:20 | |
| *** judd7_ has quit IRC | 15:20 | |
| openstackgerrit | A change was merged to openstack/swift: Allow pre-1970 dates in If-[Un]Modified-Since https://review.openstack.org/80734 | 15:20 |
| *** Guest94414 is now known as jamie_h | 15:21 | |
| portante | clayg: regarding gluster-swift, there is a bin/gluster-swift-print-metadata which prints the swift metadata contents, but nothing extensive like swift-obj-info, since an ls -l on the full object name also works | 15:22 |
| portante | anticw: we are working collecting the data for that, we have stood up a 3 node cluster (p/a/c/o), 10 ge front/back nics, 48 GB of memory each node, 12 1TB disks on each, gonna fill to 60%, time the auditor, install this change, time the auditor | 15:24 |
| *** gyee has joined #openstack-swift | 15:25 | |
| *** byeager_ has quit IRC | 15:26 | |
| *** byeager has joined #openstack-swift | 15:26 | |
| *** krtaylor has quit IRC | 15:27 | |
| *** shakamunyi has quit IRC | 15:29 | |
| *** byeager has quit IRC | 15:30 | |
| creiht | clayg: ping | 15:31 |
| creiht | I saw your comment about the pep8 stuff, but then it passed pep8 in Jenkins? | 15:31 |
| creiht | oh nm | 15:32 |
| creiht | I see what you did | 15:32 |
| creiht | silly pep8 :/ | 15:32 |
| creiht | clayg: thanks | 15:32 |
| portante | wow, clayg helps out with out saying a word, awesome | 15:39 |
| portante | ;) | 15:39 |
| portante | they call it pep8, but it is really hacking, too, right? | 15:40 |
| portante | creiht: | 15:40 |
| *** tdasilva has left #openstack-swift | 15:44 | |
| openstackgerrit | A change was merged to openstack/python-swiftclient: Use six.StringIO instead of StringIO.StringIO https://review.openstack.org/74705 | 15:47 |
| creiht | portante: heh yeah | 15:52 |
| *** byeager has joined #openstack-swift | 15:57 | |
| creiht | yay openstack politics :/ | 15:59 |
| *** byeager has quit IRC | 16:04 | |
| *** shakamunyi has joined #openstack-swift | 16:06 | |
| *** nshaikh has quit IRC | 16:10 | |
| *** shakamunyi has quit IRC | 16:10 | |
| *** piyush1 has quit IRC | 16:16 | |
| openstackgerrit | paul luse proposed a change to openstack/swift: Add Storage Policy Support to Account HEAD https://review.openstack.org/73747 | 16:22 |
| openstackgerrit | paul luse proposed a change to openstack/swift: Container Confict on obj update with wrong Storage-Policy-Index https://review.openstack.org/79731 | 16:22 |
| *** d89 has quit IRC | 16:25 | |
| notmyname | good morning | 16:29 |
| portante | mornin' | 16:30 |
| notmyname | wow. clayg was busy last night. I guess I shouldn't expect him in the office too soon ;-) | 16:30 |
| notmyname | creiht: I think we're the last group of people to complain about poor names for things ;-) | 16:30 |
| *** byeager has joined #openstack-swift | 16:31 | |
| creiht | heh | 16:32 |
| *** byeager has quit IRC | 16:35 | |
| *** piyush1 has joined #openstack-swift | 16:38 | |
| *** piyush1 has quit IRC | 16:44 | |
| peluse | notmyname: yeah and hats off to you clayg, some nice work on the policy patches last night! | 16:55 |
| *** changbl has quit IRC | 17:02 | |
| *** mkollaro has quit IRC | 17:04 | |
| *** mkollaro has joined #openstack-swift | 17:22 | |
| *** krtaylor has joined #openstack-swift | 17:27 | |
| *** changbl has joined #openstack-swift | 17:34 | |
| *** booi has quit IRC | 17:37 | |
| *** Trixboxer has quit IRC | 17:44 | |
| *** mlipchuk has quit IRC | 17:49 | |
| *** zaitcev has joined #openstack-swift | 17:51 | |
| *** ChanServ sets mode: +v zaitcev | 17:51 | |
| *** mmcardle has quit IRC | 17:53 | |
| *** mkollaro has quit IRC | 17:55 | |
| *** dmsimard1 is now known as dmsimard | 17:59 | |
| *** mkollaro has joined #openstack-swift | 18:05 | |
| *** nacim has quit IRC | 18:06 | |
| *** jamie_h has quit IRC | 18:06 | |
| *** jamie_h has joined #openstack-swift | 18:14 | |
| *** Gooch has joined #openstack-swift | 18:27 | |
| Gooch | Evening all | 18:27 |
| Gooch | Would anyone be able to tell me where on the server running the object server service the SQLLite database(s) are stored? | 18:27 |
| Gooch | We have our storage nodes running the object server service with SSD boot drives for the OS and spinning disks for the data storage | 18:28 |
| Gooch | I'd like to find out if the SQLLite DBs are stored on the SSD or the HDDs :) | 18:28 |
| glange | Gooch: objects don't have DBs, just containers and accounts | 18:30 |
| portante | Gooch: hmm, can you start by looking at your config file for "devices"? | 18:30 |
| portante | what does that say, and then check the contents of that directory | 18:30 |
| portante | df it as well to see where it lives | 18:31 |
| glange | Gooch: you could also run a command called swift-get-nodes | 18:31 |
| Gooch | Sorry, I'm not the most experianced engineer in this department | 18:31 |
| glange | no problem | 18:32 |
| Gooch | We have a container listing slow down issue due to the number of objects in certain containers | 18:32 |
| glange | do you know how many objects in the slow container? | 18:32 |
| Gooch | I'm trying to work out if currently the DB that holds the container listings is on spinning disk or SSD | 18:32 |
| Gooch | 1M+ | 18:32 |
| glange | see what happens if you run this on a server with swift and your rings: swift-get-nodes /some/path/container.ring.gz account container | 18:34 |
| Gooch | Do you have an example output for that command? | 18:34 |
| glange | http://paste.openstack.org/show/73931/ | 18:35 |
| glange | that is what it looks like on my swift all in one | 18:35 |
| Gooch | Thanks for the output :) How would this help me determin the location of object server DBs? | 18:36 |
| *** mkollaro has quit IRC | 18:36 | |
| glange | objects don't have dbs | 18:37 |
| glange | only accounts/containers | 18:37 |
| zaitcev | dude there's an ls command right there | 18:37 |
| zaitcev | scroll to the bottom | 18:37 |
| Gooch | I guess I've got an incorrect term stuck in my head. What's the name of the service that stores a database of all objects in a container? | 18:37 |
| zaitcev | glange: he's just confused. he needs a container DB | 18:37 |
| glange | zaitcev: ok | 18:38 |
| glange | Gooch: that is a container server | 18:38 |
| Gooch | Righto, that sounds like the term I needed. | 18:38 |
| glange | another place you might look is your container server configs | 18:38 |
| zaitcev | but I think it should be easy to find out where devices for a container servers are, even without swift-get-nodes. | 18:38 |
| Gooch | Do you know the default configuration for where the container server stores its DBs? | 18:39 |
| glange | zaitcev: yeah, I was telling him that in case he wanted to look for a specifc container db eventually | 18:39 |
| glange | Gooch: look in /etc/swift maybe ? | 18:39 |
| zaitcev | run swift-ring-builder /etc/swift/container.builder and it shows device names, right? Those are mounted under /srv/node, right? Then ssh and check what they are. | 18:39 |
| Gooch | My question can be kept super simple to avoid my infuriating you guys with my lack of knowledge :) | 18:40 |
| glange | haha | 18:40 |
| zaitcev | nobody is infuriated, I think | 18:40 |
| zaitcev | Ask away! | 18:40 |
| Gooch | I just need to know, on a storage server that runs all services bar the proxy, if container server DBs are stored on the OS drive or the data drives | 18:40 |
| Gooch | any help would be really appreciated. | 18:41 |
| glange | it can be configured to put the db's wherever | 18:41 |
| zaitcev | I would 1) ssh to that server, 2) look at devices= in /etc/swift/container.conf (usually /srv/node), 3) df -k and see what's mounted where | 18:41 |
| Gooch | I see. Do you know which configuration option is used to change that? | 18:41 |
| glange | do what zaitcev is saying :) | 18:42 |
| Gooch | Ok | 18:42 |
| glange | but in the container conf, it's devices = something | 18:42 |
| zaitcev | if it's omited, it defaults to /srv/nodes | 18:42 |
| zaitcev | er. /srv/node | 18:42 |
| Gooch | spinning disks are in the format /srv/node/sd*1 | 18:43 |
| Gooch | We have one SSD for boot at sda and eight spinning disks for storage at /srv/node/sd*1 | 18:43 |
| zaitcev | sounds proper | 18:44 |
| Gooch | Does that information help you help me find where the container server DBs are? :) | 18:44 |
| zaitcev | the are probably in each of these | 18:44 |
| Gooch | in /srv/node/sd* ? | 18:45 |
| zaitcev | run swift-ring-builder /etc/swift/container.builder, it should be in some safe place on an admin workstation, but some people put them on proxies... so wherever the builder files are at your place | 18:45 |
| zaitcev | the output is like | 18:45 |
| zaitcev | [root@rhev-a24c-01 swift]# swift-ring-builder container.builder | head -5 | 18:46 |
| zaitcev | ..... | 18:46 |
| zaitcev | 0 1 1 192.168.50.2 6001 192.168.50.2 6001 b1 100.00 87381 -0.00 | 18:46 |
| zaitcev | there's going to be a bunch of them | 18:46 |
| zaitcev | this one is b1 at 192.168.50.2:6001 | 18:47 |
| zaitcev | in your case should be sdb, sdc, etc. | 18:47 |
| Gooch | Yep, I follow | 18:47 |
| Gooch | What do I do with that information? | 18:47 |
| zaitcev | these device names are virtual in the sense they are mount point names. see in my case b1 is not even on /dev/b1 :-) it's just a label | 18:48 |
| zaitcev | so you need to find physical device | 18:48 |
| zaitcev | although usually when folks use your naming scheme they try to keep /dev/sdb mounted to /srv/node/sdb, but make sure some joker didn't swp them on you | 18:49 |
| Gooch | I see | 18:49 |
| Gooch | I don't need to track down the individual DB for each container | 18:49 |
| Gooch | I just need an answer on the high level question of "are they on the SSD or on the spinning disk" | 18:50 |
| *** byeager has joined #openstack-swift | 18:50 | |
| Gooch | Much of our environment is quite "default", it's not tuned as much as I'd personally like to see | 18:50 |
| zaitcev | so in my case on 192.168.50.2 it's like | 18:50 |
| zaitcev | /dev/sdb 585775624 1081832 584693792 1% /srv/node/b1 | 18:50 |
| zaitcev | at that point I just need to discover what kind of media /dev/sdb is | 18:51 |
| zaitcev | hmm | 18:51 |
| Gooch | anything in /srv/node is spinning disk. / is SSD | 18:51 |
| Gooch | So if you're telling me that the DBs live within /srv/node then they're on spinning disks :) | 18:52 |
| zaitcev | yes | 18:52 |
| zaitcev | well, I was trying to remember how to look at /proc/scsi | 18:52 |
| zaitcev | but in your case there's no alternative | 18:52 |
| Gooch | So are we, based on that info, now sure the SQLLite DBs that store container listings are stored on spinning disks? | 18:53 |
| zaitcev | okay, so for an excercise I cheated and used dmesg on my node | 18:53 |
| zaitcev | sd 4:0:1:0: [sdb] 1172123568 512-byte logical blocks: (600 GB/558 GiB) | 18:53 |
| zaitcev | looking for 4:0:1:0 | 18:53 |
| zaitcev | scsi 4:0:1:0: Direct-Access WD WD6001BKHG D1S2 PQ: 0 ANSI: 6 | 18:54 |
| zaitcev | looks like a WD Black | 18:54 |
| zaitcev | which is spinning | 18:54 |
| zaitcev | Case Closed! | 18:54 |
| Gooch | So the next question then | 18:54 |
| Gooch | How do I move the container server and all it's SQLLite DBs from spinning disk to SSD? :) | 18:55 |
| Gooch | (Hint: Change all my storage nodes to SSD only is not an answer I can afford :) ) | 18:55 |
| zaitcev | I would buy a number of SSDs that match replication + handout number (swift-ring-builder will tell what it is for you, but usually 3 + 1). | 18:56 |
| Gooch | zaitcev: I have to go now for a break. I really appreciate your help | 18:56 |
| zaitcev | Attach them to nodes that will be hosting containers, fdisk, etc | 18:56 |
| Gooch | I'll be back later tonight or maybe tommorow for sure | 18:56 |
| zaitcev | okay | 18:57 |
| Gooch | Thanks again dude(tte) | 18:57 |
| ahale | i'd probably just makdir /srv/node/ssd, not have anything mounted , disable mount check for account/container and then migrate containers over to them slowly with new rings | 18:57 |
| zaitcev | it's an intriguing possibility if root is large enough | 18:58 |
| zaitcev | we'll poke him for ls and du outputs later, I'm curious how big those DBs are actually | 18:59 |
| ahale | yeah true, it does need a decent size root, but can always add ssd later on the mounts if it gets a problem | 18:59 |
| *** byeager_ has joined #openstack-swift | 19:06 | |
| *** byeager has quit IRC | 19:09 | |
| openstackgerrit | A change was merged to openstack/swift: Enable object versions for SAIO by default https://review.openstack.org/81615 | 19:19 |
| openstackgerrit | A change was merged to openstack/swift: Fix race on container recreate https://review.openstack.org/81104 | 19:19 |
| clayg | acoles-: gawd paste.openstack is slow | 19:28 |
| peluse | clayg/torgomatic: both acct HEAD and 409 patches are fully rebased and ready for business | 19:35 |
| * clayg really needs to look at the acct HEAD patch :\ | 19:37 | |
| clayg | but it's so big! | 19:37 |
| clayg | can we just punt on that? | 19:37 |
| *** tdasilva has joined #openstack-swift | 19:45 | |
| *** vu has joined #openstack-swift | 19:46 | |
| openstackgerrit | Clay Gerrard proposed a change to openstack/swift: additional functional tests for account acls https://review.openstack.org/80073 | 19:55 |
| *** shri has joined #openstack-swift | 19:59 | |
| *** nacim has joined #openstack-swift | 20:03 | |
| openstackgerrit | Clay Gerrard proposed a change to openstack/swift: Clean up Account ACL sysmeta in AccountController https://review.openstack.org/80140 | 20:05 |
| *** tusharsg has joined #openstack-swift | 20:05 | |
| openstackgerrit | Clay Gerrard proposed a change to openstack/swift: Harden v2 common.middleware.acl interface https://review.openstack.org/80188 | 20:11 |
| clayg | is the ring reload time configurable? | 20:24 |
| clayg | like I know it's paramaterized when you create a Ring - but like is that plumbed through a conf anywhere? | 20:24 |
| creiht | clayg: doesn't look like it is plummed | 20:27 |
| creiht | would be easy to add though | 20:28 |
| clayg | i've never needed it, maybe it's fine | 20:28 |
| clayg | what's the default like - 30? | 20:29 |
| creiht | 15 | 20:29 |
| clayg | sweet | 20:29 |
| peluse | clayg: jsut commented in github. yeah, it would need to be plumbed but nobody requires it now and it would be easy to add | 20:37 |
| peluse | and since the def is 15 I didn't bother to add the plumbing at this point in time | 20:38 |
| peluse | clayg: the acct HEAD patch is kinda big in line count but its pretty simple and more of a small set of large chunks of code as opposed to a big pile of little changes if that helps :) | 20:39 |
| clayg | no it doesn't help - but I am still going to need to review it i suppose | 20:41 |
| peluse | I tried to punt on it too but notmyname said no - we need a way to provide/report per policy usage for billing... | 20:44 |
| *** jamie_h has quit IRC | 20:44 | |
| clayg | heheh - i was kidding | 20:45 |
| clayg | but it should *obviously* be X-Storage-Policy-Bronze-Object-Count to mirror X-Account-Object-Count rather than X-Storage-Policy-Object-Count-Bronze :D | 20:46 |
| peluse | I actually had it both ways at one point or another believe it or not and settled with adding the policy at the end :) | 20:50 |
| notmyname | back from meetings. should be here most of the rest of the day now | 20:52 |
| creiht | will the default policy show up as X-Account-Object-Count? | 20:53 |
| notmyname | creiht: no, the current headers will continue to mean the same thing | 20:54 |
| torgomatic | -Count and -Bytes-Used are sums of everything | 20:54 |
| notmyname | ie the sum of everything | 20:54 |
| creiht | ahh ok | 20:54 |
| creiht | well that's still fine | 20:54 |
| notmyname | creiht: we knew you wouldn't like that, so just didn't even try :-) | 20:54 |
| creiht | just want to make sure we will not have to update billing systems with the rollout | 20:54 |
| creiht | hehe | 20:55 |
| notmyname | creiht: oh yeah. we've all got billing systems to update | 20:55 |
| *** krtaylor has quit IRC | 20:57 | |
| peluse | creiht: here's the patch if you're interested: https://review.openstack.org/#/c/73747/10 | 20:57 |
| * peluse is anxious to see clayg's review of some of his creative python in the acct HEAD patch | 20:58 | |
| *** nacim has quit IRC | 21:00 | |
| *** tdasilva has left #openstack-swift | 21:01 | |
| creiht | peluse: thx | 21:01 |
| *** JuanManuelOlle has quit IRC | 21:07 | |
| *** tongli has quit IRC | 21:07 | |
| *** greghaynes has quit IRC | 21:11 | |
| clayg | peluse: you're python is fine | 21:13 |
| clayg | peluse: that patch is trixy | 21:13 |
| creiht | man clayg, you have been busy | 21:15 |
| notmyname | I just got an email about accepted talks to the summit. I'm assuming others of you should also get some | 21:15 |
| * creiht is saving his good stuff to get a ticket to paris :) | 21:16 | |
| notmyname | :-) | 21:16 |
| creiht | or I might have just been lazy | 21:16 |
| peluse | notmyname: indeed | 21:24 |
| peluse | notmyname: on the QA cluster thing though, Intel has a break-out room and we can present there | 21:25 |
| notmyname | peluse: ok. I also made a summit session for ti | 21:28 |
| *** vu has quit IRC | 21:32 | |
| *** byeager_ has quit IRC | 22:01 | |
| *** byeager has joined #openstack-swift | 22:01 | |
| *** byeager has quit IRC | 22:05 | |
| *** dmsimard has quit IRC | 22:08 | |
| *** byeager has joined #openstack-swift | 22:14 | |
| *** sungju_ has joined #openstack-swift | 22:23 | |
| *** sungju- has joined #openstack-swift | 22:25 | |
| peluse | clayg: I'm heading out for the night - thanks a million for the feedback thus far on the rollup thing, will address those things in the morning (updated patch I mean, comments already posted)... | 22:27 |
| *** sungju_ has quit IRC | 22:27 | |
| *** byeager has quit IRC | 22:28 | |
| *** byeager has joined #openstack-swift | 22:28 | |
| *** jergerber has quit IRC | 22:30 | |
| *** tburnes has quit IRC | 22:30 | |
| *** tburnes has joined #openstack-swift | 22:31 | |
| notmyname | can we have a "swift 2048"? track->sector->block->drives->inode->(X)FS->server->zone->DC->region->cluster | 22:31 |
| *** byeager has quit IRC | 22:33 | |
| *** changbl has quit IRC | 22:47 | |
| *** tusharsg has quit IRC | 23:32 | |
| openstackgerrit | Samuel Merritt proposed a change to openstack/swift: Purge self.policies from Swift https://review.openstack.org/81940 | 23:34 |
Generated by irclog2html.py 2.14.0 by Marius Gedminas - find it at mg.pov.lt!