| *** kragniz_ is now known as kragniz | 00:00 | |
| notmyname | mattoliverau: jrichli_ I just found something interesting that is probably pretty small that can be looking into | 00:14 |
|---|---|---|
| notmyname | *looked | 00:14 |
| jrichli_ | notmyname: sounds great, what is it? | 00:15 |
| notmyname | a weird, and I think extraneous we're doing | 00:15 |
| notmyname | so reset your SAIO, then get an auth token from tempauth | 00:15 |
| notmyname | that's it | 00:15 |
| notmyname | now look at your logs | 00:15 |
| notmyname | you'll see one like this: | 00:15 |
| notmyname | Feb 16 00:10:31 saio swift_proxy: - - 16/Feb/2015/00/10/31 HEAD /v1/v1.0 HTTP/1.0 204 - Swift - - - - tx88f2881a8c994748aeb84-0054e13577 Host:%20saio:8080%0AUser-Agent:%20Swift%0AX-Trans-Id:%20tx88f2881a8c994748aeb84-0054e13577 0.0151 GET_INFO - 1424045431.786942005 1424045431.802002907 - | 00:16 |
| notmyname | so there's 2 things there | 00:16 |
| notmyname | first, the GET_INFO swift source isn't documented on http://docs.openstack.org/developer/swift/logs.html | 00:16 |
| notmyname | but, more interesting is why are we making a HEAD request to "/v1/v1.0"???! | 00:16 |
| notmyname | that's the odd part | 00:17 |
| jrichli_ | ok, I will take a look. | 00:17 |
| notmyname | it's line 524 in proxy/controllers/base.py that's making the call | 00:17 |
| notmyname | but I don't know why the path is set and we're doing that | 00:17 |
| notmyname | jrichli_: thanks :-) | 00:17 |
| notmyname | I've got to get offline for a bit (I think my family is starting to ask why I'm on my computer this afternoon :-) | 00:18 |
| jrichli_ | np. I will let you know what I find out. | 00:18 |
| notmyname | but, for anyone else (acoles_away, daisuke when he gets online, slogging is updated to work with current swift) | 00:18 |
| mattoliverau | notmyname: I'll let daisuke now when he turns up | 00:19 |
| notmyname | mattoliverau: thanks | 00:19 |
| mattoliverau | sorry in a meeting | 00:19 |
| notmyname | https://github.com/notmyname/slogging #link | 00:19 |
| mattoliverau | ta | 00:19 |
| notmyname | be back later :-) | 00:20 |
| *** ho has joined #openstack-swift | 00:43 | |
| ho | good morning guys! | 00:44 |
| notmyname | jrichli_: did you find anything? | 00:44 |
| notmyname | ho: how was your trip home? | 00:44 |
| ho | notmyname: good. I have no jet lag! | 00:45 |
| notmyname | that's great! | 00:45 |
| ho | notmyname: thanks! | 00:46 |
| jrichli_ | notmyname: I had to do some things. I haven't gotten far yet in the investigating. | 00:46 |
| notmyname | jrichli_: no worries | 00:46 |
| *** ho has quit IRC | 00:47 | |
| *** ho has joined #openstack-swift | 00:51 | |
| notmyname | seems that anything after the auth_prefix in the path is getting treated like a swift account and a HEAD is done | 00:52 |
| jrichli_ | seems like it's doing the HEAD in order to set the info in the environment and in memcache (because the info was cached) | 01:13 |
| jrichli_ | *not cached | 01:13 |
| jrichli_ | ah, and you are saying the info wasn't found in cache because the wrong path was used? | 01:14 |
| openstackgerrit | John Dickinson proposed openstack/swift: added swift_source to ratelimit info calls https://review.openstack.org/156095 | 01:15 |
| notmyname | jrichli_: that's one issue ^ | 01:15 |
| notmyname | jrichli_: the swift_source wasn't set on a call to get_account_info in the ratelimit middleware | 01:15 |
| jrichli_ | i saw that had to have been None for GET_INFO to be used | 01:16 |
| notmyname | right | 01:16 |
| notmyname | hmm...so here's the situation now | 01:17 |
| notmyname | it's not tempauth (although that's where I saw it first) | 01:17 |
| notmyname | any request to `http://saio:8080/foo/bar` will cause "Feb 16 01:17:15 saio swift_proxy: - - 16/Feb/2015/01/17/15 HEAD /v1/bar HTTP/1.0 204 - Swift - - - - txbdb842e7fec64662b79d5-0054e1451b Host:%20saio:8080%0AUser-Agent:%20Swift%0AX-Trans-Id:%20txbdb842e7fec64662b79d5-0054e1451b 0.0091 RL - 1424049435.027575016 1424049435.036645889 -" | 01:18 |
| notmyname | along with the associated account server requests to up to 9 different servers (3 + 6 handoffs) | 01:18 |
| notmyname | which is the real issue | 01:18 |
| notmyname | which is coming from the ratelimit middleware (the "RL" -- with my patch) | 01:20 |
| jrichli_ | I see. Makes sense. How is the swift_source chosen? I mean, RL sounds good for something called ratelimit, but does each calling class just use any string? | 01:23 |
| notmyname | jrichli_: basically :-) | 01:23 |
| jrichli_ | ok | 01:23 |
| notmyname | jrichli_: along with the "please consult and add yours to http://docs.openstack.org/developer/swift/ratelimit.html" | 01:23 |
| notmyname | I mean http://docs.openstack.org/developer/swift/logs.html | 01:23 |
| jrichli_ | oh, yes. good to know! | 01:25 |
| jrichli_ | I will add your fix and run again. Just maybe I will give you a +1! | 01:27 |
| notmyname | thanks. | 01:32 |
| * notmyname off to cook dinner | 01:32 | |
| *** kajinamit has joined #openstack-swift | 01:34 | |
| *** kajinamit has quit IRC | 01:36 | |
| openstackgerrit | Hisashi Osanai proposed openstack/swift: Enable Object Replicator's failure count in recon https://review.openstack.org/138342 | 01:47 |
| *** jamielennox is now known as jamielennox|away | 02:19 | |
| *** jamielennox|away is now known as jamielennox | 02:27 | |
| *** doxavore has joined #openstack-swift | 02:43 | |
| *** doxavore has quit IRC | 02:46 | |
| *** doxavore has joined #openstack-swift | 02:46 | |
| *** ahonda has quit IRC | 02:47 | |
| *** sluo_wfh has quit IRC | 02:49 | |
| *** sluo_wfh has joined #openstack-swift | 02:50 | |
| *** sluo_wfh has quit IRC | 02:55 | |
| *** sluo_wfh has joined #openstack-swift | 02:56 | |
| *** doxavore has quit IRC | 03:45 | |
| *** wer_ has quit IRC | 04:12 | |
| *** wer has quit IRC | 04:12 | |
| *** wer has joined #openstack-swift | 04:12 | |
| *** wer_ has joined #openstack-swift | 04:12 | |
| *** BAKfr has quit IRC | 04:32 | |
| *** BAKfr has joined #openstack-swift | 04:32 | |
| *** ppai has joined #openstack-swift | 04:35 | |
| *** BAKfr has quit IRC | 04:37 | |
| *** SkyRocknRoll has joined #openstack-swift | 04:44 | |
| *** SkyRocknRoll has joined #openstack-swift | 04:46 | |
| *** BAKfr has joined #openstack-swift | 04:48 | |
| *** SkyRocknRoll has quit IRC | 04:52 | |
| torgomatic | notmyname: we could use a pastedeploy composite to route auth down a different popeline than non-auth | 05:19 |
| torgomatic | pipeline, even | 05:19 |
| torgomatic | although now I'm wondering what a popeline is | 05:20 |
| notmyname | torgomatic: that's not a bad idea | 05:20 |
| notmyname | torgomatic: it probably involves some sort of smoke | 05:20 |
| notmyname | the response codes are color-coded | 05:20 |
| notmyname | you have to be patient, though. latency is a killer | 05:20 |
| notmyname | torgomatic: is this sane? Request\.blank\('/[^(v1)]/ | 05:21 |
| notmyname | ie to match Request.blank('/a/c/o | 05:21 |
| notmyname | 'cause I wanna change it to Request.blank('/v1/a/c/o | 05:22 |
| torgomatic | notmyname: Request.blank with a regex pattern? I don't think I would know what that means | 05:24 |
| torgomatic | oh wait | 05:24 |
| torgomatic | yeah, that's probably okay | 05:25 |
| notmyname | torgomatic: I got it. actually jrichli_ got it, but still | 05:25 |
| notmyname | it wasn't ok actually | 05:25 |
| notmyname | Request\.blank\('/[^(v1)]*/ | 05:25 |
| torgomatic | other than the fact you're parsing Python with a regex, but neglecting that, it's okay | 05:25 |
| notmyname | heh | 05:25 |
| torgomatic | "Other than that, Mrs. Lincoln, how was the play?" | 05:25 |
| notmyname | actually no. just for grep or my editor to find stuff in the tests | 05:25 |
| torgomatic | ah, fair enough :) | 05:25 |
| *** jrichli_ has quit IRC | 05:36 | |
| notmyname | wow. this test is relying on a bug (in the test) in order to pass | 05:42 |
| mattoliverau | Test... Relying on a bug.. A test! Wow | 05:57 |
| notmyname | anyone seen or know anything about keystone's authenticated tokens? from the ML it seems they're faster than uuid tokens | 05:59 |
| notmyname | and bounded in size | 06:00 |
| openstackgerrit | Merged openstack/swift: added swift_source to ratelimit info calls https://review.openstack.org/156095 | 06:09 |
| *** kota_ has joined #openstack-swift | 06:10 | |
| notmyname | kota_: how was your trip home? | 06:10 |
| kota_ | ya, it was good. I got very good health in Japan. | 06:11 |
| notmyname | great! glad you made it ok | 06:11 |
| kota_ | Thanks! How was the dinner at the last day of hackathon? | 06:11 |
| mattoliverau | Delicious :) we went out for Thai | 06:13 |
| notmyname | great! the best part was meeting joearnold on top of a hill in the city with his binoculars and looking at jupiter, orion, the pliedes (subaru cluster), and even the andromeda galexy | 06:13 |
| mattoliverau | +1 to urban astronomy :) | 06:13 |
| notmyname | https://twitter.com/joearnold/status/566101434232340480 | 06:13 |
| notmyname | it was pretty cool | 06:13 |
| kota_ | Wow, so beautiful! | 06:14 |
| notmyname | torgomatic: mattoliverau: I (briefly) looked at adding a test for swift_source to that ratelimit patch that just added. looks like there's some others that test that (with FakeSwift). but test_ratelimit doesn't use FakeSwift, so getting that plumbed through is a little more difficult | 06:15 |
| *** ahonda has joined #openstack-swift | 06:15 | |
| mattoliverau | But we got it in 1 patchset.. Don't break the record :p | 06:21 |
| notmyname | heh | 06:22 |
| notmyname | torgomatic: the reason I updated the logs doc on the fist patchset is because I'm the one who originally had to troll through the code to write that doc int he first place. the pain isn't easily forgotten ;-) | 06:25 |
| openstackgerrit | OpenStack Proposal Bot proposed openstack/swift: Imported Translations from Transifex https://review.openstack.org/155967 | 06:30 |
| ho | notmyname: AE tokens? thanks for the info. I will read the spec. BTW does swift have a problem like performance with keystone so far? | 06:40 |
| notmyname | ho: generally swift doesn't have the performance problem ;-) | 06:40 |
| ho | notmyname: i see. | 06:41 |
| notmyname | ho: but yea, keystone has historically struggled with how to manage a large number of keys. swift, in normal operation, can use a huge number of keys, so keystone becomes a problem | 06:42 |
| *** kota_ has quit IRC | 06:42 | |
| notmyname | ho: and the PKI tokens have the problem with being unbounded in size. and in general just being really big (multiple KB). and that makes things really hard when you have to send that for every swift request. especially since PKI tokens aren't cacheble | 06:42 |
| ho | notmyname: i think so too. I have been developing policy based RBAC with keystone. if there are a lot of tokens, roles in wire increase. | 06:44 |
| notmyname | so all of that is why I'm pretty interested in the AE tokens they talked about. | 06:45 |
| *** kota_ has joined #openstack-swift | 06:48 | |
| ho | notmyname: yeah. thanks for the info again! | 06:50 |
| notmyname | here it is https://review.openstack.org/#/c/130050/ | 06:50 |
| *** kota_ has quit IRC | 06:57 | |
| notmyname | good night, all | 06:59 |
| ho | notmyname: good night! | 07:00 |
| *** SkyRocknRoll has joined #openstack-swift | 07:20 | |
| *** SkyRocknRoll has joined #openstack-swift | 07:20 | |
| *** SkyRocknRoll has quit IRC | 07:21 | |
| *** chlong has quit IRC | 07:51 | |
| *** nellysmitt has joined #openstack-swift | 07:53 | |
| openstackgerrit | Merged openstack/swift: Imported Translations from Transifex https://review.openstack.org/155967 | 07:58 |
| *** jistr has joined #openstack-swift | 08:08 | |
| *** rledisez has joined #openstack-swift | 08:12 | |
| *** joeljwright has joined #openstack-swift | 08:12 | |
| *** silor has joined #openstack-swift | 08:26 | |
| *** jordanP has joined #openstack-swift | 08:42 | |
| *** jordanP has quit IRC | 08:44 | |
| *** jordanP has joined #openstack-swift | 08:44 | |
| *** openstackgerrit has quit IRC | 08:57 | |
| *** openstackgerrit has joined #openstack-swift | 08:57 | |
| *** jistr has quit IRC | 08:59 | |
| *** mmcardle has joined #openstack-swift | 09:01 | |
| *** madhuri_ has quit IRC | 09:05 | |
| donagh | /server Aquarius.panicbnc.net 1337 donagh/freenode:2fe6CDf8 | 09:09 |
| *** madhuri_ has joined #openstack-swift | 09:12 | |
| *** ho has quit IRC | 09:14 | |
| *** madhuri_ has quit IRC | 09:31 | |
| *** madhuri_ has joined #openstack-swift | 09:32 | |
| *** madhuri_ has quit IRC | 09:37 | |
| remix_tj | donagh: ehm. | 09:52 |
| openstackgerrit | Merged openstack/swift: Add functional tests for container TempURLs https://review.openstack.org/155513 | 09:53 |
| *** sluo_wfh has quit IRC | 10:04 | |
| *** ppai has quit IRC | 10:27 | |
| *** ppai has joined #openstack-swift | 10:41 | |
| *** aix has joined #openstack-swift | 10:44 | |
| *** miqui_ has quit IRC | 11:18 | |
| *** miqui has joined #openstack-swift | 11:18 | |
| *** anticw has quit IRC | 11:19 | |
| *** anticw has joined #openstack-swift | 11:20 | |
| *** aix has quit IRC | 11:27 | |
| *** mwilliams_ct has joined #openstack-swift | 11:35 | |
| *** mwilliams_ct has left #openstack-swift | 11:37 | |
| *** ppai has quit IRC | 12:31 | |
| *** aix has joined #openstack-swift | 12:32 | |
| *** silor has quit IRC | 12:34 | |
| *** silor has joined #openstack-swift | 12:35 | |
| *** ppai has joined #openstack-swift | 12:44 | |
| *** mahatic has joined #openstack-swift | 12:46 | |
| *** EmilienM is now known as EmilienM|afk | 13:20 | |
| *** silor has quit IRC | 13:28 | |
| *** silor has joined #openstack-swift | 13:29 | |
| *** silor1 has joined #openstack-swift | 13:37 | |
| *** silor has quit IRC | 13:40 | |
| *** sandywalsh_ has quit IRC | 13:52 | |
| *** tdasilva has joined #openstack-swift | 13:59 | |
| *** acoles_away is now known as acoles | 14:29 | |
| *** doxavore has joined #openstack-swift | 14:30 | |
| doxavore | I had another quick question on replica placement - this weekend I learned that Swift distributes evenly among failure domains of a given tier, so if I had 3 regions, each would get 1 replica. Does this happen even if the weight of the regions aren't equal? | 14:33 |
| doxavore | I'm thinking of a cluster where drives/nodes are periodically needing to be removed - it seems if it takes total weight of each region into account, then it'd be nearly impossible to ensure you have 1 replica of everything in each region? | 14:34 |
| *** ppai has quit IRC | 14:52 | |
| *** jrichli_ has joined #openstack-swift | 15:01 | |
| *** sweeper has joined #openstack-swift | 15:05 | |
| sweeper | doxavore: it uses a weighted-nearest neighbor round robin to distribute availability on a cyclomatic basis | 15:06 |
| *** EmilienM|afk is now known as EmilienM | 15:16 | |
| *** sandywalsh has joined #openstack-swift | 15:47 | |
| *** dmsimard_away is now known as dmsimard | 15:47 | |
| *** haomaiwang has joined #openstack-swift | 15:47 | |
| *** nellysmitt has quit IRC | 15:51 | |
| *** abhirc has joined #openstack-swift | 15:54 | |
| *** jistr has joined #openstack-swift | 15:58 | |
| *** david-lyle_afk is now known as david-lyle | 16:01 | |
| *** rdaly2 has joined #openstack-swift | 16:02 | |
| *** jrichli_ has quit IRC | 16:09 | |
| *** jrichli_ has joined #openstack-swift | 16:10 | |
| *** zigo has quit IRC | 16:26 | |
| *** delattec has joined #openstack-swift | 16:30 | |
| *** zigo has joined #openstack-swift | 16:31 | |
| *** jistr has quit IRC | 16:32 | |
| *** delattec has quit IRC | 16:33 | |
| *** delatte has quit IRC | 16:33 | |
| notmyname | good morning | 16:40 |
| notmyname | donagh: might want to change your password now | 16:41 |
| donagh | notmyname: Yeah, I noticed right after I did it. Thanks for pointing it out (also thanks cschede: for same) | 16:43 |
| notmyname | donagh: we missed you last week | 16:43 |
| donagh | I was tracking indirecty. Reports from Alistair and Gerry were very good. | 16:44 |
| donagh | s/were/are/ | 16:44 |
| acoles | mattoliverau: donagh exists ^^ :D | 16:45 |
| *** wer has quit IRC | 16:45 | |
| *** wer_ has quit IRC | 16:45 | |
| donagh | acoles: I guess that's in in joke! | 16:45 |
| acoles | donagh: see hipchat | 16:46 |
| donagh | Ah | 16:46 |
| acoles | notmyname: i am missing the sfo climate | 16:46 |
| *** wer_ has joined #openstack-swift | 16:47 | |
| notmyname | it was warm this weekend. about 70-75 (F) | 16:47 |
| *** IRTermite has quit IRC | 16:47 | |
| *** wer has joined #openstack-swift | 16:47 | |
| notmyname | mattoliverau: look at acoles talking to himself! | 16:47 |
| acoles | notmyname: lol | 16:49 |
| notmyname | acoles: donagh: do you know anything about the authenticated encryption tokens that keystone is recently talking about? | 16:49 |
| acoles | notmyname: not yet | 16:49 |
| acoles | notmyname: do you have any link? | 16:50 |
| notmyname | http://dolphm.com/benchmarking-openstack-keystone-token-formats/ and https://review.openstack.org/#/c/130050/ is what I've seen | 16:50 |
| acoles | thx | 16:52 |
| donagh | notmyname: dito (i.e., no) | 16:54 |
| *** silor1 has quit IRC | 17:05 | |
| acoles | clayg: tried devstack install. failed :( http://paste.openstack.org/show/175328/ will keep banging head against wall | 17:05 |
| doxavore | notmyname: when you mentioned this weekend that swift distributes replicas "evenly" across failure domains, did you mean with or without weights? it seems it's be difficult to keep 1 copy in 3 different regions (or zones) unless they were _exactly_ the same total weight if that's how works, no? | 17:18 |
| doxavore | as long as my domains were roughly the same size, it would be likely that _most_ partitions existed in all 3 regions, but if it's expecting them to be perfectly divided, but that doesn't really satisfy a requirement on making sure there's a copy of everything :-/ | 17:22 |
| *** mahatic has quit IRC | 17:25 | |
| *** tellesnobrega has quit IRC | 17:25 | |
| *** arnaud_o has quit IRC | 17:25 | |
| *** MooingLemur has quit IRC | 17:25 | |
| *** mtreinish has quit IRC | 17:25 | |
| *** Anticimex has quit IRC | 17:25 | |
| *** joearnold has quit IRC | 17:25 | |
| *** JoshNang has quit IRC | 17:25 | |
| *** klrmn has quit IRC | 17:25 | |
| *** bobby2 has quit IRC | 17:25 | |
| *** notmyname has quit IRC | 17:25 | |
| *** eikke has quit IRC | 17:25 | |
| *** mikehn_ has quit IRC | 17:25 | |
| *** portante has quit IRC | 17:25 | |
| *** ahale has quit IRC | 17:25 | |
| *** rdaly2 has quit IRC | 17:25 | |
| *** rdaly2 has joined #openstack-swift | 17:25 | |
| *** rledisez has quit IRC | 17:28 | |
| *** notmyname has joined #openstack-swift | 17:31 | |
| *** ChanServ sets mode: +v notmyname | 17:31 | |
| *** mahatic has joined #openstack-swift | 17:31 | |
| *** tellesnobrega has joined #openstack-swift | 17:31 | |
| *** arnaud_o has joined #openstack-swift | 17:31 | |
| *** klrmn has joined #openstack-swift | 17:31 | |
| *** MooingLemur has joined #openstack-swift | 17:31 | |
| *** mtreinish has joined #openstack-swift | 17:31 | |
| *** Anticimex has joined #openstack-swift | 17:31 | |
| *** joearnold has joined #openstack-swift | 17:31 | |
| *** JoshNang has joined #openstack-swift | 17:31 | |
| *** bobby2 has joined #openstack-swift | 17:31 | |
| *** eikke has joined #openstack-swift | 17:31 | |
| *** mikehn_ has joined #openstack-swift | 17:31 | |
| *** portante has joined #openstack-swift | 17:31 | |
| *** ahale has joined #openstack-swift | 17:31 | |
| mahatic | I see. I just read the wiki page on it and checked back here, your explanation is clearly crisp :D | 17:31 |
| notmyname | mahatic: and now it just fixed itself | 17:32 |
| mahatic | notmyname, I see. I just read the wiki page on it and checked back here, your explanation is clearly crisp :D | 17:32 |
| mahatic | yeah :D | 17:32 |
| openstackgerrit | John Dickinson proposed openstack/swift-specs: added clarification on how to ask questions https://review.openstack.org/156291 | 17:33 |
| notmyname | tdasilva: ^^ there | 17:34 |
| tdasilva | notmyname: nice, thanks! | 17:36 |
| tdasilva | notmyname, acoles: yes, really missing SFO climate. woke up to coldest morning in 11 years here with -35F for wind chill | 17:38 |
| *** acoles is now known as acoles_away | 17:38 | |
| notmyname | wow!! | 17:38 |
| *** mkerrin has quit IRC | 17:39 | |
| *** nellysmitt has joined #openstack-swift | 17:40 | |
| notmyname | mahatic: how's your patch progress coming? | 17:41 |
| mahatic | notmyname, I wrote a test and it's throwing up an error (trying to fix it). Will push that if I don't get to resolve. | 17:43 |
| notmyname | ok, great | 17:44 |
| notmyname | mahatic: I'll be out (my) thursday and friday this week. so if you have questions, please let me know sooner than later | 17:44 |
| *** abhirc has quit IRC | 17:44 | |
| mahatic | notmyname, yeah sure. I remember that | 17:44 |
| notmyname | :-) | 17:45 |
| notmyname | FYI. last friday I had a really interesting phone call with a company that is going to open-source a tape-library connector for Swift that they've written | 17:49 |
| notmyname | I had several conversations with companies in Paris about this topic, so I know there are other people out there looking to do something similar | 17:49 |
| remix_tj | notmyname: you say using swift as VTL? | 17:49 |
| notmyname | so if you or people at your company are involved in this in any way, please get in touch with me | 17:50 |
| notmyname | remix_tj: no, not exactly. using swift to store data on tape rather than magnetic memory. so they've taken swift's DiskFile abstraction and written one that works with their VTL. I don't know yet how generic it is or if there are any proprietary hooks (I think it's pretty general, though) | 17:51 |
| notmyname | remix_tj: so you can imagine a deployment that has a tape policy and a drive policy and a flash policy, all in different regions | 17:51 |
| remix_tj | oh, so you say using a tape drive as tier | 17:52 |
| *** briancline has quit IRC | 17:52 | |
| *** briancline has joined #openstack-swift | 17:52 | |
| notmyname | briancline: TIL that .zone is a valid TLD | 17:54 |
| *** mmcardle has quit IRC | 17:54 | |
| *** acoles_away is now known as acoles | 18:01 | |
| *** rdaly2 has quit IRC | 18:02 | |
| *** reed has joined #openstack-swift | 18:02 | |
| *** rdaly2 has joined #openstack-swift | 18:03 | |
| *** EmilienM is now known as EmilienM|afk | 18:07 | |
| acoles | tdasilva: ouch (-35) | 18:14 |
| *** rdaly2 has quit IRC | 18:22 | |
| tdasilva | acoles: yeah, that's a bit colder than SFO :-) | 18:24 |
| *** rdaly2 has joined #openstack-swift | 18:25 | |
| tdasilva | I wrote a hackathon trip report to be sent internally and ended up converting to a blog post, Here's the link in case anyone is interested: https://thiagodasilvablog.wordpress.com/2015/02/16/openstack-swift-mid-cycle-report/ | 18:26 |
| notmyname | tdasilva: I'm planning on going skiing next weekend. but I'm worried that there won't be enough snow | 18:26 |
| notmyname | tdasilva: thanks for the link | 18:26 |
| tdasilva | Also feel free to send me corrections and I'll be glad to update | 18:27 |
| tdasilva | notmyname: come over to NE, we got plenty of snow here :P | 18:27 |
| notmyname | tdasilva: that's much more detailed than my blog post I rushed out on friday morning! :-) | 18:27 |
| notmyname | tdasilva: you have Intel listed twice in the first paragraph | 18:29 |
| jrichli_ | tdasilva: Thanks for sharing the write up. It is a great summary! | 18:29 |
| notmyname | it is a great summary | 18:29 |
| acoles | tdasilva: yeah, i was just wondering if there had been a merger with intel i hadn't heard about ;) | 18:29 |
| notmyname | ya, he 2nd Intel is HP | 18:30 |
| tdasilva | oops! thanks! | 18:30 |
| notmyname | acoles: surprise! that company split everyone is talking about? now it's intel :-) | 18:30 |
| acoles | tdasilva: great write up! thanks | 18:31 |
| *** Trixboxer has joined #openstack-swift | 18:32 | |
| notmyname | tdasilva: and the appropriate twitter accounts I have access to have now mentioned it :-) | 18:34 |
| tdasilva | acoles, jrichli_, notmyname: thanks! | 18:34 |
| acoles | jrichli_: guess what, my day is gone and i didn't get round to the encryption spec updates but its on the agenda for tomorrow | 18:36 |
| jrichli_ | acoles: np. I have been playing catch up all day as well. I hope to have some code to push in a few days. We shall see. | 18:37 |
| *** MasterPiece has joined #openstack-swift | 18:38 | |
| *** jordanP has quit IRC | 18:40 | |
| *** rdaly2_ has joined #openstack-swift | 18:43 | |
| *** rdaly2 has quit IRC | 18:43 | |
| *** IRTermite has joined #openstack-swift | 18:55 | |
| *** IRTermite has quit IRC | 18:57 | |
| *** IRTermite has joined #openstack-swift | 18:59 | |
| *** IRTermite has quit IRC | 19:01 | |
| *** rdaly2 has joined #openstack-swift | 19:03 | |
| *** rdaly2_ has quit IRC | 19:04 | |
| *** IRTermite has joined #openstack-swift | 19:07 | |
| *** acoles is now known as acoles_away | 19:11 | |
| *** aix has quit IRC | 19:11 | |
| *** IRTermite has quit IRC | 19:15 | |
| *** rdaly2_ has joined #openstack-swift | 19:23 | |
| *** rdaly2 has quit IRC | 19:23 | |
| sweeper | ok, so this ring deal. if I have two physical regions, with one zone in region B and 2 zones in region A, and I set a replication level of 3, will I end up with the expected distribution of replications (one per zone, 1-B1, 2-A1, 3-A2)? | 19:41 |
| notmyname | sweeper: are the zones the same size? | 19:42 |
| sweeper | notmyname: I'd prefer not to have to care about that | 19:42 |
| notmyname | hmm...actually I don't think that matters | 19:43 |
| *** rdaly2_ has quit IRC | 19:43 | |
| notmyname | I think you have to care about relative sizes. swift cannot evenly distribute data across failure domains if those failure domains don't match in capacity | 19:44 |
| *** rdaly2 has joined #openstack-swift | 19:44 | |
| notmyname | ...but let's come back to that later | 19:44 |
| notmyname | sweeper: what will happen in you scenario is that you'll end up with one replica in on region, and 2 replicas in the other region until the smaller region is filled up. then you'll end with 3 replicas in the large region for the remainder of the capacity available | 19:45 |
| notmyname | I think. torgomatic, is that right? | 19:45 |
| notmyname | or cschwede? | 19:45 |
| torgomatic | eh... kind of? | 19:45 |
| sweeper | notmyname: mkay, what if the zone sizes are ever (or perhaps I fudge using overload?) | 19:46 |
| sweeper | s/ever/even/ | 19:46 |
| torgomatic | if your smaller region is exactly 1/2 of the larger region, then it's 1/3 of the total capacity, and you get each partition with 1 replica in the smaller and 2 in the larger | 19:46 |
| torgomatic | if it's slightly more than 1/3 of the total capacity, then you get most partitions with 1 replica in the smaller and 2 in the larger | 19:46 |
| torgomatic | the remainer have 2 in the smaller and 1 in the larger | 19:47 |
| torgomatic | *remainder | 19:47 |
| sweeper | ok. is there a simpler way to do this? basically I have a datacenter where we use the data, and an offsite backup location, I need two copies at the DC and one offsit | 19:47 |
| notmyname | torgomatic: ah. thanks for the correction | 19:47 |
| sweeper | it's not really acceptable to have even a few partitions with 2 replicas in the smaller region | 19:49 |
| *** MasterPiece has quit IRC | 19:49 | |
| sweeper | torgomatic: what if we overloaded the larger region? | 19:50 |
| torgomatic | sweeper: well, overload is a global setting, not per-device | 19:50 |
| torgomatic | also I don't think it'd help | 19:50 |
| torgomatic | if your smaller region were less than 1/3 of the total, then normally you'd get some parts with all 3 replicas in the larger | 19:51 |
| torgomatic | setting an overload would alleviate that | 19:51 |
| *** abhirc has joined #openstack-swift | 19:51 | |
| sweeper | torgomatic: so a slightly smaller backup region plus overload, gives me what I want? | 19:53 |
| torgomatic | sweeper: I believe so | 19:53 |
| sweeper | awesome, thanks | 19:53 |
| torgomatic | of course, your smaller region will fill faster than the larger since its disks will be proportionally fuller | 19:53 |
| torgomatic | so watch out for that | 19:53 |
| sweeper | but it should only need to be asymptotically smaller, to force the ring builder to distribute the right way? | 19:54 |
| sweeper | e.g. I start with 20gb in region A and 9gb in region B, then I can go to 200gb in region A and 99gb in region B, while maintaining the same availability profile? | 19:56 |
| torgomatic | sweeper: yes, but you'll probably want a little padding for if you lose some disks in the larger region too, since you don't ever want 2 replicas of any partition in the smaller | 19:57 |
| torgomatic | so like 100 disks / 50 disks is probably bad; you'd want more like 100 disks / 45 disks so you can lose a few in the larger region without upsetting the delicate balance | 19:57 |
| torgomatic | that's all I'm saying | 19:57 |
| sweeper | torgomatic: hmm. so are we just doing this at too small of a scale? it seems like a reasonable first architecture, but one that is pretty complex to effect | 19:58 |
| torgomatic | sweeper: well, you're sort of fighting the ring builder here; Swift's ring builder was really designed with the idea of having a bunch of zones that data could be evenly spread across | 20:00 |
| torgomatic | paying attention to disk weights (sizes), of course | 20:00 |
| torgomatic | it's much more about just standing up a bunch of storage in various failure domains and letting data spread around as far as it can | 20:01 |
| torgomatic | something like you're doing, where you're forcing 2 replicas on the left and 1 on the right, is sort of fighting against what the ring builder wants | 20:01 |
| torgomatic | it's possible, but it's a little fragile | 20:01 |
| torgomatic | that said, if you want to do this with only a few disks, you can | 20:02 |
| sweeper | so would 3 zones be a better fit then? | 20:02 |
| torgomatic | just maintain the property that sum(big-region-weights) >= 2 * sum(little-region-weights) | 20:02 |
| sweeper | I mean, basically we have 2 physical failure domains, that we have to spread across, and within one physical failure domain we have two network/power domains | 20:03 |
| *** rdaly2_ has joined #openstack-swift | 20:03 | |
| torgomatic | so if big-region loses a disk and you want to carve it out of the rings, you may have to fudge the other weights upwards to keep that invariant | 20:03 |
| torgomatic | sweeper: yeah, if you did 3 zones and set overload=999999999999999, you'd get 1 replica per zone | 20:04 |
| *** rdaly2 has quit IRC | 20:04 | |
| *** EmilienM|afk is now known as EmilienM | 20:04 | |
| *** abhirc has quit IRC | 20:05 | |
| *** rdaly2_ has quit IRC | 20:06 | |
| *** rdaly2 has joined #openstack-swift | 20:06 | |
| sweeper | torgomatic: I almost feel like there should be layers of rings? I mean I get the notion of evenly distributing data, but it feels like the ring builder throws out any notions of intra-zone availability in favor of global availability? | 20:07 |
| torgomatic | sweeper: yeah, the assumption is that all regions are equal, and all zones within a region are equal, and all nodes, etc etc | 20:09 |
| torgomatic | this sort of primary-and-backup scenario is one of the places where that falls down | 20:09 |
| sweeper | well even something like "dc in us and smaller dc in au" seems like it would be problematic | 20:10 |
| torgomatic | well, only if you add in the requirement of 2 replicas in .us and 1 in .au | 20:11 |
| torgomatic | if you let the replicas fall where they may, then it works out okay | 20:11 |
| sweeper | yea I suppose | 20:12 |
| sweeper | hmm. I guess one could just write one's own ring-building logic... | 20:12 |
| torgomatic | true; once the ring-builder writes out the .ring.gz file, it's just a static mapping | 20:12 |
| torgomatic | the running Swift code knows nothing of how the partitions got that way, just that there's this big fat array to index into | 20:13 |
| sweeper | aight. well I don't particularly feel led to start mucking around in internals at this point, but I'm pretty happy it's not too scary an enterprise | 20:22 |
| sweeper | torgomatic: thanks for the help \o | 20:22 |
| torgomatic | no problem :) | 20:36 |
| *** rdaly2 has quit IRC | 20:43 | |
| *** mahatic has quit IRC | 20:49 | |
| *** mahatic has joined #openstack-swift | 20:57 | |
| *** Nadeem_ has joined #openstack-swift | 20:58 | |
| *** devlaps has joined #openstack-swift | 21:01 | |
| mattoliverau | Morning all | 21:09 |
| notmyname | hi mattoliverau | 21:09 |
| mattoliverau | acoles_away: I'm impressed with your ability to use 2 irc clients at once, very impressed :P | 21:09 |
| notmyname | lol | 21:09 |
| tdasilva | mattoliverau: hey! good morning | 21:10 |
| mattoliverau | tdasilva: Morning, so your not frozen (just read scroll back)! I'm about to read your blog post :) | 21:10 |
| tdasilva | mattoliverau: not frozen, but you did remind me of a Frozen (the disney movie) meme I saw the other day | 21:16 |
| jrichli_ | mattoliverau: good morning | 21:17 |
| mattoliverau | tdasilva: you can't just dangle the mention of a meme and not share it :P | 21:18 |
| tdasilva | mattoliverau: looking for it | 21:18 |
| mattoliverau | jrichli_: Morning! You came back again, we obviously aren't as scary as we thought | 21:18 |
| jrichli_ | so far, so good :-) | 21:20 |
| *** ChanServ sets mode: +v mattoliverau | 21:25 | |
| tdasilva | mattoliverau: some good ones here: https://www.facebook.com/MassachusettsMemes | 21:26 |
| panbalag | hi.I'm trying to configure swift as the backend for Glance. The glance.conf has a parameter 'swift_store_key'. Any idea where I can find this? | 21:32 |
| mattoliverau | tdasilva: lol, if it makes you feel any better its going to be only 86 here today, so a cooler day :P | 21:33 |
| notmyname | panbalag: is that something to do with auth credentials so glance can talk to swift? | 21:33 |
| panbalag | notmyname, yeah | 21:34 |
| notmyname | panbalag: ok. I'd guess it's something set up by or in keystone then | 21:34 |
| panbalag | notmyname, ok. let me try that | 21:34 |
| notmyname | tdasilva: can you give me a rhel/cent version of https://launchpad.net/ubuntu/+source/swift | 21:37 |
| notmyname | hmm...why don't we have rsync in the requirements file? | 21:40 |
| notmyname | I guess since it's not a python thing? | 21:52 |
| doxavore | are there any recommendations for swift 2.2.2 installation on ubuntu? it doesn't appear anyone has released packages and the documentation points to Ubuntu Cloud Archive still at 2.2.0. (2.2.2 was the first time that overload setting was made available I take it?) | 21:55 |
| notmyname | doxavore: installation isn't hard. or rather, there's nothing special. but if you aren't building your own packages, you'll be dependent on whatever your distro gives you | 21:57 |
| notmyname | maybe I'm splitting hairs there | 21:57 |
| openstackgerrit | John Dickinson proposed openstack/swift: update the getting started doc https://review.openstack.org/156388 | 22:00 |
| *** os1 has joined #openstack-swift | 22:01 | |
| notmyname | doxavore: most deployers I know do build and manage their own packages of swift so that they can stay up to date. the distros normally only update a couple of times a year | 22:01 |
| os1 | Hi | 22:01 |
| os1 | Is there a reason why 'swift-init all status' would return that container-server and object-server are not running, | 22:01 |
| notmyname | os1: do you reason to believe that they are running? | 22:02 |
| os1 | but doing a 'ps aux | grep swift' reveals processes for both the container-server and object-server? | 22:02 |
| notmyname | ah ok :-) | 22:02 |
| notmyname | os1: probably the pid file was lost. | 22:02 |
| *** aix has joined #openstack-swift | 22:02 | |
| os1 | notmyname : Could you be more specific? | 22:02 |
| *** jamielennox is now known as jamielennox|away | 22:03 | |
| os1 | notmyname : Doing a 'swift-init {container-server | object-server} restart', then, should work, but it fails whenever trying to bring up one of those services. | 22:05 |
| notmyname | os1: the process id (pid) is written down in a file, but if that file is lost or overwritten, then swift-init doesn't know where to look to query a process. | 22:05 |
| notmyname | os1: why does it fail? config or maybe because there's already something bound on the port | 22:05 |
| notmyname | os1: you might be able to use `swift-orphans` or `swift-oldies` to help you find old processes that need to be cleaned up | 22:07 |
| notmyname | os1: and if you keep having trouble starting those processes, then either start them directly or look in the logs to see if there's something it's reporting | 22:07 |
| os1 | notmyname : Okay. By 'start them directly', which utility are you referring to? | 22:11 |
| os1 | Thank you. | 22:11 |
| *** jamielennox|away is now known as jamielennox | 22:14 | |
| notmyname | os1: `swift-init` is a tool that has some aliases and finds some config files. but ultimately, you can use eg `swift-proxy-server /etc/swift/proxy-server.conf -v` | 22:14 |
| os1 | notmyname : I see. Thank you. | 22:16 |
| notmyname | anyone else going to the operators summit in Philadelphia next month? | 22:18 |
| notmyname | err...wow. it's already the middle of february. so like in a few weeks | 22:18 |
| *** abhirc has joined #openstack-swift | 22:22 | |
| mattoliverau | I think I'm travelled out.. until vancouver at least :P | 22:23 |
| tdasilva | notmyname: do you mean swift packages for rhel/centos ? | 22:29 |
| tdasilva | notmyname: I assuming your are mentioning RDO, which you can find it here: https://repos.fedorapeople.org/repos/openstack/ | 22:30 |
| *** nellysmitt has quit IRC | 22:37 | |
| notmyname | tdasilva: thanks. it has to do with https://review.openstack.org/#/c/156388/ | 22:38 |
| *** nellysmitt has joined #openstack-swift | 22:38 | |
| notmyname | tdasilva: I linked https://openstack.redhat.com/Repositories which seems fairly general | 22:38 |
| *** nellysmitt has quit IRC | 22:42 | |
| tdasilva | notmyname: cool, just added some comments there | 22:45 |
| notmyname | thanks | 22:45 |
| notmyname | tdasilva: what? you didn't know you work for shuttleworth now? ;-) | 22:46 |
| tdasilva | lol | 22:46 |
| tdasilva | I rode in an elevator with him in Paris and he didn't say anything | 22:46 |
| tdasilva | :-) | 22:46 |
| tdasilva | notmyname: ubuntu already released packages for swift 2.2.2 huh? | 22:47 |
| notmyname | seems that way | 22:48 |
| notmyname | I only found that out myself today when I clicked that link | 22:48 |
| StevenK | swift | 2.2.2-0ubuntu1 | vivid | source | 22:48 |
| StevenK | Yeah | 22:48 |
| openstackgerrit | John Dickinson proposed openstack/swift: update the getting started doc https://review.openstack.org/156388 | 22:48 |
| notmyname | tdasilva: done | 22:49 |
| *** fbo has quit IRC | 22:49 | |
| *** fbo has joined #openstack-swift | 22:54 | |
| mattoliverau | phew, expenses report done, that wasn't a waste of a morning at all :P | 22:55 |
| notmyname | :-) | 22:58 |
| notmyname | so I just saw this job posting on twitter: IBM looking for someone to run Swift. http://searchamericanjobs.com/display-job/29898533/SoftLayer-OpenStack-Swift-Lead-Architect.html | 22:59 |
| notmyname | that's cool, but they're looking for someone with 5+ years experience running Swift at scale | 22:59 |
| notmyname | hmmm | 22:59 |
| mattoliverau | lol | 23:00 |
| jrichli_ | interesting | 23:00 |
| notmyname | so they are targeting about 4 of the ops guys at rackspace? who themselves only have about 4.5 years experience with swift in prod at scale :-) | 23:00 |
| notmyname | (counting from may 2010) | 23:00 |
| mahatic | I did this first git pull origin master; it showed me I have an updated file and would I want it to be merged, I said yes. This is what git log looked like - >http://paste.openstack.org/show/175531/ | 23:03 |
| mahatic | my today's changes are committed with git commit --amend (which amended the wrong commit - the merge one). I then did a git reset --soft HEAD~1 | 23:05 |
| mahatic | I now have a ton of files to committed | 23:05 |
| mahatic | wrong channel to post this? :) | 23:05 |
| notmyname | mahatic: no, it's fine. you've got problems with swift, so let's get it sorted | 23:06 |
| notmyname | (but I'm going to go pour myself some coffee first) | 23:06 |
| mahatic | great, thanks! | 23:06 |
| mahatic | sure | 23:06 |
| *** Nadeem_ has quit IRC | 23:07 | |
| *** chlong has joined #openstack-swift | 23:09 | |
| mahatic | git status shows this -> http://paste.openstack.org/show/175532/ | 23:14 |
| *** doxavore has quit IRC | 23:14 | |
| notmyname | mahatic: I'm back | 23:14 |
| mahatic | yay | 23:14 |
| notmyname | do you have changes locally that you're trying to keep? | 23:14 |
| notmyname | ah I see | 23:14 |
| notmyname | you amended the merge commit | 23:14 |
| mahatic | yeah! | 23:15 |
| notmyname | but you need to undo that and amend the right one | 23:15 |
| mahatic | and then i did this git reset --soft HEAD~1 | 23:15 |
| notmyname | have you pushed anything (I'm guessing no since I didn't see a bot message in here) | 23:15 |
| mahatic | no I didn't | 23:15 |
| notmyname | ok, that's good | 23:15 |
| notmyname | how big was your change? | 23:16 |
| notmyname | or, how sad would you be if you had to redo it? | 23:16 |
| mahatic | I do have a back up. | 23:16 |
| notmyname | ah good | 23:16 |
| mahatic | so nothing really sad :) | 23:16 |
| notmyname | :-) | 23:16 |
| notmyname | ok, then `git reset --hard origin/master` will get you back to the consistent view of what's upstream | 23:16 |
| mahatic | okay, on it | 23:17 |
| *** rdaly2 has joined #openstack-swift | 23:18 | |
| notmyname | mahatic: I've changed my git workflow slightly so that I can avoid weird stuff like that | 23:18 |
| notmyname | mahatic: instead of using `git pull`, I use `git fetch --all`. that pulls down changes, but it doesn't update my local branches or the working directory | 23:19 |
| notmyname | mahatic: then, if everything is ok, I can easily `git merge origin/master` and be just fine | 23:19 |
| mahatic | notmyname, oh okay. I will note this | 23:19 |
| mahatic | but now, how do I amend my existing commit? | 23:19 |
| notmyname | so splitting the pull into the fetch+merge basically gives me a nice place where I can eg track your changes without having to affect my own branches | 23:20 |
| *** joeljwright has quit IRC | 23:20 | |
| notmyname | mahatic: is the commit that you want to amend on its own brach? | 23:20 |
| notmyname | branch | 23:20 |
| mahatic | nope, master | 23:21 |
| *** jrichli_ has quit IRC | 23:21 | |
| notmyname | well that's not good. but we can fix that | 23:21 |
| notmyname | is this amending a patch you have already submitted upstream? | 23:21 |
| *** jrichli has joined #openstack-swift | 23:21 | |
| mahatic | notmyname, https://review.openstack.org/#/c/153617/ it's not in the upstream yet | 23:22 |
| notmyname | ok | 23:22 |
| notmyname | mahatic: so we need to do a little surgery on your git repo so that things keep working. | 23:23 |
| mahatic | I see | 23:23 |
| notmyname | here's what's happened: you wrote a change on master and submitted it, but it hasn't landed before other patches landed upstream. so now the upstream repo is out of sync with your local master branch. and that's why the `git pull` tried to do the merge | 23:24 |
| notmyname | make sense? | 23:24 |
| mahatic | yup | 23:24 |
| mahatic | so I did merge | 23:25 |
| notmyname | again? | 23:25 |
| mahatic | no, this merge was a couple of days back | 23:25 |
| notmyname | oh | 23:25 |
| mahatic | what you said happened when I tried to pull a couple of days back (and I merged) | 23:26 |
| notmyname | ok | 23:26 |
| mahatic | and now, today after my changes I did this git commit --amend (so the prev merge commit was amended) | 23:26 |
| notmyname | got it | 23:26 |
| mahatic | okay | 23:26 |
| notmyname | try to keep your local master clean. that is, the only time you update it is when you merge from origin. which should always be a fast-forward merge. all of your work happens on your local branches | 23:27 |
| mahatic | oh okay. I should create a local branch for my local changes | 23:27 |
| notmyname | yes | 23:28 |
| notmyname | and in this case, to fix this issue, it will be really easy | 23:28 |
| notmyname | since you've already pushed it to gerrit | 23:28 |
| notmyname | the first version. you'll still have to redo your updates to it | 23:28 |
| mahatic | okay | 23:28 |
| notmyname | sound ok? ready to get started on that? | 23:28 |
| mahatic | yeah, sounds good | 23:28 |
| mahatic | ready! | 23:29 |
| notmyname | mahatic: ok. you already did the git reset? everything is back to normal on master? | 23:29 |
| mahatic | yeah, that's right | 23:29 |
| mahatic | git reset --hard origin/master | 23:29 |
| mahatic | did that | 23:30 |
| notmyname | ok | 23:30 |
| notmyname | ok, now we'll ask git-review to download the patch set and it will create the branch for you: `git review -d 153617` | 23:30 |
| notmyname | (that's instead of rewinding and doing all the git stuff locally. this way you just pull the branch from something external. much easier) | 23:31 |
| notmyname | that will give you some weirdly-named branch | 23:31 |
| notmyname | ie it downloads that patch into a branch and checks out that branch | 23:31 |
| mahatic | okay | 23:31 |
| mahatic | that's right. this sounds a bit familiar (I really need to get a hang of this!) | 23:31 |
| notmyname | what branch did it create? | 23:31 |
| notmyname | do `git branch` and look for the one with a * next to it | 23:32 |
| mahatic | it's still downloading | 23:32 |
| notmyname | oh | 23:32 |
| mahatic | * review/mahati/153617 | 23:32 |
| mahatic | so that one | 23:32 |
| notmyname | ok | 23:33 |
| mahatic | my changes are back! | 23:33 |
| notmyname | so now, let's rename it something that's easier for you to deal with (I'm guessing you don't want to type `git checkount review/mahati/153617` every time you need to work on it | 23:33 |
| notmyname | to rename the current branch, do this: `git branch -m <new name here>` | 23:34 |
| mahatic | sure | 23:34 |
| notmyname | so maybe "checker" or something. whatever you want. | 23:34 |
| *** rdaly2 has quit IRC | 23:34 | |
| notmyname | mahatic: what OS are you using? | 23:34 |
| mahatic | shouldn't I be giving a generic one? So that going forward I can use this same local branch for all my other changes? | 23:35 |
| mahatic | fedora | 23:35 |
| notmyname | no. you'll have a new branch for every patch | 23:35 |
| mahatic | fedora 20 | 23:35 |
| notmyname | and you can safely delete it after it's landed | 23:35 |
| mahatic | ah okay | 23:35 |
| mahatic | I should make this a practice | 23:36 |
| notmyname | so there's a tool called gitk which might help you understand what's going on with git. it shows the git commit history visually. | 23:36 |
| mahatic | * ringchecker (that's my branch) | 23:37 |
| mahatic | oh | 23:37 |
| notmyname | on the mac, I use gitx. it's basically the same thing. here's a screenshot: http://d.not.mn/gitx.png | 23:37 |
| mahatic | nice | 23:37 |
| notmyname | so you can see that right now I've got a update_docs, fix_get_info, and master branches locally. and also the origin and gerrit remotes | 23:37 |
| notmyname | ok, so one more thing before you redo your changes | 23:38 |
| notmyname | now that you want to make updates to your ringchecker branch, you need to make sure it is still ok with the current state of master. ie there won't be any merge conflicts when it lands | 23:38 |
| notmyname | make sense? | 23:38 |
| mahatic | yeah | 23:39 |
| mahatic | and how do i do that? | 23:39 |
| notmyname | easy | 23:39 |
| notmyname | you're on your branch | 23:39 |
| notmyname | so `git rebase master`. that will replay the commits on your current branch (ringchecker) on top of the master branch. in your case, you've only got one patch | 23:40 |
| notmyname | it will warn if there are conflicts. if so, then you'll need to resolve those and I can help you with that | 23:40 |
| notmyname | but normally it will go without any errors | 23:40 |
| mahatic | oh okay | 23:41 |
| mahatic | on it | 23:41 |
| mahatic | and this step can be done after I redo my changes too? | 23:41 |
| notmyname | it's a lot easier if you do the rebase first | 23:42 |
| mahatic | as in, is this a step that needs to be taken every time I make changes on my local branch? | 23:42 |
| notmyname | you only really need to do the rebase if there are merge conflicts in getting it onto master | 23:43 |
| *** tdasilva has quit IRC | 23:43 | |
| mahatic | oh okay | 23:43 |
| mahatic | done | 23:43 |
| mahatic | no conlicts | 23:43 |
| mahatic | conflicts | 23:43 |
| notmyname | great | 23:45 |
| notmyname | so now put your patches onto teh current state of the ringchecker branch | 23:45 |
| notmyname | and then `git commit --amend -a` to add them to the existing commit | 23:45 |
| mahatic | my earlier question was: say for my next patch, i create a new branch, should i have to run that rebase master command? | 23:45 |
| notmyname | then `git review` for all the rest of us to se | 23:45 |
| mahatic | :) | 23:45 |
| notmyname | mahatic: in general no | 23:46 |
| mahatic | it should happen when I commit, right | 23:46 |
| mahatic | ? | 23:46 |
| notmyname | what will? | 23:46 |
| mahatic | when I submit for review and get approval, the merge to master happens | 23:46 |
| mahatic | sorry, when I do my next new patch and commit, it should be on master (without me running git rebase), correct? | 23:47 |
| notmyname | correct | 23:48 |
| notmyname | wait | 23:48 |
| notmyname | correct to "when I submit for review and get approval, the merge to master happens" | 23:48 |
| notmyname | when you do another patch later, you'll create a new branch for it (git checkout -b cool_feature) and then do your work there | 23:48 |
| mahatic | yes and then my cool_feature does need an explicit rebase by git rebase master? | 23:49 |
| notmyname | mahatic: no, not when you create it. you'll create it off of master | 23:50 |
| notmyname | mahatic: you'll only need to do a rebase if your patch hasn't landed and there have been conflicting changes on master | 23:50 |
| notmyname | summit conference session voting has started! https://www.openstack.org/vote-vancouver/ | 23:51 |
| mahatic | ah okay | 23:51 |
| notmyname | looks like a lot of good swift talks proposed | 23:51 |
| *** ChanServ changes topic to "Summit voting: https://www.openstack.org/vote-vancouver/ | riority Reviews: https://wiki.openstack.org/wiki/Swift/PriorityReviews | Review Dashboard: http://goo.gl/r2mxbe | Ideas: https://wiki.openstack.org/wiki/Swift/ideas" | 23:53 | |
| notmyname | link is in the topic now | 23:53 |
| *** treyd has joined #openstack-swift | 23:56 | |
| notmyname | mahatic: and the last important point is that when you do the amend, do not change the change-id line. that ties it all together in gerrit | 23:56 |
| mahatic | notmyname, sure. I actually never amend that. I thought it was generated automatically | 23:57 |
| mahatic | amended | 23:57 |
| notmyname | it is via a commit hook you have locally. git review set that up for you at the beginning when you did `git review -s` | 23:57 |
| *** jkugel has quit IRC | 23:58 | |
| mahatic | ah okay | 23:58 |
| openstackgerrit | Mahati proposed openstack/swift: Ring checker in swift-recon https://review.openstack.org/153617 | 23:59 |
| mahatic | finally there it is! | 23:59 |
| mahatic | thanks so much notmyname ! | 23:59 |
Generated by irclog2html.py 2.14.0 by Marius Gedminas - find it at mg.pov.lt!
