Tuesday, 2010-09-21

*** ppetraki has joined #openstack00:00
*** metcalfc has quit IRC00:04
*** metcalfc has joined #openstack00:04
*** gustavomzw has quit IRC00:10
*** scottie has joined #openstack00:17
*** lmcdowell has quit IRC00:22
*** jeromatron has joined #openstack00:33
jeromatronanyone know if the Compute project is planning on providing the capability to have multiple IO devices?00:36
*** tobym has joined #openstack00:36
*** miclorb has quit IRC00:37
*** localhost has quit IRC00:45
*** localhost has joined #openstack00:47
*** tobym has quit IRC00:47
*** localhost has quit IRC00:48
*** localhost has joined #openstack00:54
*** burris has quit IRC00:57
*** miclorb_ has joined #openstack00:59
*** joearnol_ has quit IRC01:06
*** ppetraki has quit IRC01:07
*** mtaylor has quit IRC01:16
*** tobym has joined #openstack01:21
*** maplebed has quit IRC01:25
*** ArdRigh has joined #openstack01:28
*** burris has joined #openstack01:30
*** jeromatron has quit IRC01:33
*** mtaylor has joined #openstack01:43
*** ChanServ sets mode: +v mtaylor01:43
*** sirp has quit IRC01:46
*** sirp1 has joined #openstack01:46
*** jeromatron has joined #openstack01:48
*** ar1 has joined #openstack01:58
*** scottie has quit IRC02:25
*** lmcdowell has joined #openstack02:31
*** miclorb_ has quit IRC02:33
*** lmcdowell has quit IRC02:43
*** mtaylor has quit IRC02:47
*** sophiap has quit IRC02:47
*** maple_be1 has quit IRC02:55
*** metoikos has quit IRC02:56
*** metoikos has joined #openstack02:57
*** sophiap has joined #openstack02:58
*** miclorb has joined #openstack03:11
*** metoikos has quit IRC03:18
*** metoikos has joined #openstack03:19
*** xtoddx has joined #openstack03:27
*** sophiap has quit IRC03:28
*** kevnfx has joined #openstack03:43
uvirtbotNew bug: #644075 in swift "account-reaper does not have an except/log global catchall" [Undecided,New] https://launchpad.net/bugs/64407503:46
*** kashyapc has joined #openstack03:50
*** jeromatron has quit IRC03:52
*** jeromatron has joined #openstack03:53
*** cyonyx has joined #openstack04:01
*** stewart has quit IRC04:06
*** stewart has joined #openstack04:19
*** cyonyx has quit IRC04:20
*** omidhdl has joined #openstack04:22
*** kashyapc has quit IRC04:26
*** tobym has quit IRC04:40
*** sureshgv has quit IRC04:45
*** sureshgv has joined #openstack04:45
*** kashyapc has joined #openstack04:47
*** ArdRigh has quit IRC04:50
*** f4m8_ is now known as f4m804:54
*** gaveen has joined #openstack04:55
*** gaveen has joined #openstack04:55
*** joearnold has joined #openstack04:55
*** gaveen has joined #openstack04:56
*** omidhdl1 has joined #openstack05:01
uvirtbotNew bug: #644092 in nova "authorization not checked in ec2 api" [High,New] https://launchpad.net/bugs/64409205:01
*** omidhdl has quit IRC05:01
*** kevnfx has quit IRC05:04
*** abecc has quit IRC05:14
*** DubLo7 has quit IRC05:20
*** DubLo7 has joined #openstack05:22
*** omidhdl1 has left #openstack05:43
*** miclorb has quit IRC05:43
*** ibarrera has joined #openstack05:45
*** joearnold has quit IRC05:46
*** Kami__ has joined #openstack06:00
*** miclorb has joined #openstack06:11
*** sirp1 has quit IRC06:25
*** dele_ted has joined #openstack06:36
*** chmouel_ has joined #openstack06:41
*** chmouel has quit IRC06:42
*** miclorb has quit IRC06:48
*** allsystemsarego has joined #openstack06:48
*** allsystemsarego has joined #openstack06:49
*** brd_from_italy has joined #openstack07:00
*** dele_ted has quit IRC07:12
*** calavera has joined #openstack07:16
*** Kami__ has quit IRC07:25
*** Kami_ has joined #openstack07:26
*** dele_ted has joined #openstack07:34
*** zheng_li has joined #openstack07:40
*** dele_ted has quit IRC08:05
*** jtimberman has quit IRC08:34
*** jtimberman has joined #openstack08:35
*** mtaylor has joined #openstack09:12
*** ChanServ sets mode: +v mtaylor09:12
*** cw has quit IRC09:31
*** cw has joined #openstack09:31
*** chmouel_ is now known as chmouel09:36
*** DubLo7 has quit IRC09:37
*** gaveen has quit IRC10:17
*** gaveen has joined #openstack10:30
*** gaveen has quit IRC10:31
*** gustavomzw has joined #openstack10:36
*** sophiap has joined #openstack10:51
*** cloudmeat1 has joined #openstack10:58
*** cloudmeat has quit IRC11:01
*** scottie has joined #openstack11:04
*** ar1 has quit IRC11:08
*** kashyapc has quit IRC11:20
*** kashyapc has joined #openstack11:21
*** pietro has quit IRC11:21
*** ambo has quit IRC11:27
*** ambo has joined #openstack11:27
*** gustavomzw has quit IRC11:29
*** ctennis has quit IRC11:34
*** krzycoder has quit IRC11:38
*** tobym has joined #openstack11:52
*** pietro has joined #openstack12:04
*** localhost has quit IRC12:09
*** localhost has joined #openstack12:26
*** sophiap has quit IRC12:28
*** sophiap has joined #openstack12:29
*** sophiap has quit IRC12:31
*** ctennis has joined #openstack12:41
*** gholt has quit IRC12:42
*** letterj has quit IRC12:42
*** clayg has quit IRC12:42
*** dgoetz has quit IRC12:42
*** sophiap has joined #openstack12:42
*** tobym has quit IRC12:45
*** hazmat has joined #openstack12:46
*** klord has joined #openstack12:46
*** piken has joined #openstack12:50
*** kashyapc has quit IRC12:52
*** hazmat has quit IRC12:55
*** tobym has joined #openstack13:09
*** blamar has quit IRC13:11
*** ppetraki has joined #openstack13:14
*** gondoi has joined #openstack13:16
*** clayg has joined #openstack13:17
*** gholt has joined #openstack13:19
*** Podilarius has joined #openstack13:20
*** krish has joined #openstack13:20
*** dendrobates is now known as dendro-afk13:30
*** gustavomzw has joined #openstack13:37
chmouelthe /etc/sudoer.d/nova_sudoers from nova-common make sudo coredump backtrace pretty heavily! http://paste.openstack.org/show/28/13:41
*** dendro-afk is now known as dendrobates13:45
sorenchmouel: orly? Weirdness.13:53
sorenchmouel: Erk.13:53
chmouelyeah it can't parse the file nfi why13:54
*** abecc has joined #openstack13:54
chmouelgot to reboot in recoverty to get root again on that box13:55
*** matclayton has joined #openstack13:55
matclaytonjust wondering where do you set the replication factor in swift? and can it be changed on a per file basis?13:56
notmynamematclayton: it's constant right now (global to the cluster13:57
creihtmatclayton: Currently it is set when creating the ring, though there are parts that still a bit hard coded to expect 3 replicas13:57
creihtmatclayton: do you want more or fewer replicas?13:58
matclaytoncreiht: ok thanks, our use case is that some files become "hot" and therefore we need to put them on more than one machine to serve the load (or three even)13:58
matclaytonbut only for short periods13:58
creihtso that is a bit different13:58
creihtthe replicas are more for durability13:58
creihtadding more replicas will not necessarily mean faster serving13:59
creihtat least the way the code is written at the moment13:59
matclaytonessentially we run a music hosting site (lots of mp3s) and looking at using swift for storage13:59
creihta better solution to what you are talking about is to have some sort of hot cache in front of swift13:59
matclaytonis this a good solution14:00
matclaytonyeah we use to run a hot cache on ssd's and can easily redeply that14:00
creihtat RS, we typically use the CDN integration for that type of use case14:00
matclaytonso whats the reasoning behind adding more replica's not increasing throughput14:00
matclayton(we would increase the proxies as well)14:01
matclaytonon average the load is fairly predictable, its just some hot spot as content becomes popular which we have found to cause problems in the past14:01
creihtwell it isn't coded in a way to try to evenly distribute read request accross all replicas14:01
matclaytonoh does it prefer a primary one?14:02
creihtit returns whichever one it finds first14:02
matclaytonah ok14:02
creihtwhich can be a bit arbitrary14:03
creihtnot that the code couldn't be changed to try do distribute the reads more evenly14:03
matclaytonsounds to me like that might self regulate a bit14:03
matclaytonor we can patch it, we are quite keen on swift being a python shop14:03
creihtall that said, it seems like it would be better to use something in front of swift to cache the hot content14:04
matclaytonindeed, we would do that anyway14:04
*** f4m8 is now known as f4m8_14:04
matclaytonprobably just setup a couple of varnish nodes14:04
creihtwe were just talking about that :)14:04
matclaytonthe issue is our content size is average 100meg14:04
matclaytonwe have done varnish in front of S3 before and its not without issues14:05
matclaytonso holding them in memory doesn't work too well :)14:05
notmynamewhat sort of issues? it's not a setup we run, but it's common enough that you aren't the first to mention something like that14:05
matclaytonthe other question is, how performant is swift at delivering the files? what is the likely bottleneck?14:05
creihtthe bottleneck will most likely be the network14:06
notmynamenetwork to the storage nodes14:06
matclaytonbetween proxy and object?14:06
creihtreads are quite fast14:06
matclaytonis there anyway to just return a reference to files on disk and use that to serve files direct off object node through something like nginx14:06
creihtmatclayton: how many nodes are you running, and what type of networking are you using?14:07
matclaytonwe run 4 storage nodes at the moment14:07
matclaytonabout 20TB14:07
matclaytonhold 500Mbit out going14:07
matclaytongrows by about 30-40% monthly14:07
matclaytoncurrently this is just 4 machines with nginx and a dumb file system14:08
matclaytonall machines have 1Gig connection, internal and external14:08
creihtso swift should be able to saturate your 500Mb pretty easily14:09
creihtare you runing proxies on the storage nodes as well?14:09
matclaytonthats our current load, we are looking at switching to some other setup to be defined14:09
matclaytonwe can run proxies wither on the same nodes or seperate14:09
creihtavg file size is 100MB, or is that the upper end?14:10
matclaytonbasically because increases capacity is a real pain14:10
matclaytoncurrently average is 60Meg, however that is likely to increase14:10
matclaytonwe have 100M upload limit, which soon as we can, we would love to remove14:10
*** xfaf has joined #openstack14:11
*** rnirmal has joined #openstack14:11
creihtso when an object becomes hot, what type of throughput are we talking about?14:12
matclaytonso the other thought we had bounced around, was can we setup nginx on the object nodes, and give it access to the filesystem beneath swift, and get swift just to return the node_address and filename, preventing the internal traffic14:12
matclaytoncreiht: dont know the stats at the moment for that14:13
matclaytonbut we can be looking at few thousand people listening to the same file14:13
creihtmatclayton: we played with using senfile once, but if I remember correctly it didn't make that much of a difference14:13
matclaytonfor serving direct from the object node to the final clients, avoiding the proxy?14:14
creihtI'm trying to recall... it has been a while :)14:14
matclaytonknow that feeling,14:14
*** cloudmeat has joined #openstack14:15
*** cloudmeat1 has quit IRC14:16
*** xfaf has quit IRC14:16
creihtis there any way you can simulate a similar load?14:16
matclaytonwe could try14:17
matclaytonnot going to be easy, not really setup for it at the moment14:17
creihtit would be interesting to test it first to see where the weak points are14:17
notmynamethe sendfile test was to see if avoiding user space (memory) would be faster. turned out that the network was a much more limiting factor than kernel/user memory space14:17
*** kevnfx has joined #openstack14:18
matclaytonyou are talking about proxy->obj store network?14:18
notmynamewe've mentioned exposing every storage node to public traffic, but I don't know that many (any) tests have been done on that14:18
matclaytonso I was wondering if you can just remove that step14:18
notmynamein our setup, proxy -> obj network is smaller than public -> proxy net, so that becomes the limiting factor14:19
creihtmatclayton: it is just code right? :)14:19
matclaytonexactly :)14:19
creihtI can't think of anything that would prevent you from trying it14:19
matclaytonso is most of your traffic internal then?14:19
creihtWe have a good amount of both external and internal traffic14:20
creihtnot sure which is more actually :)14:20
matclaytonah ok14:20
creihtbasically we have several trunked 10G connection coming in from the outside14:20
creihteach proxy has 10G to the outside14:20
creihteach storage node has 1G to everything14:21
*** dendrobates is now known as dendro-afk14:23
matclaytoncool, will give it a go as is and see how we get on, might try the serve direct approach. out of interest was there any specific reason it wasn't done that way originally?14:23
matclaytonsuppose you need the client to be aware of the system14:23
creihtwe talked about it14:23
creihtone of the main reasons that I can remember is that it is easier to handle failure scenarios if everything is going through the proxy14:24
creihtAnd it was more than "good enough" :)14:24
matclaytonyeah makes sense,14:24
creihtwe also like that the actual storage nodes are on a private network, not accesible directly by the public networks14:24
matclaytonoh i think this is the last couple of questions. Any reason for xfs over ext4? or something we shoudl watch out for on ext4?14:24
*** kevnfx has quit IRC14:25
creihtfor testing in our use case, xfs performance degraded much less over time than ext414:25
creihtyour use case is a bit different than ours, so I can't say weather or not there is a difference14:25
creihteverything should work fine on ext414:26
*** dendro-afk is now known as dendrobates14:26
matclaytongreat thanks14:26
notmynamejust make sure that xattrs are enabled14:26
matclaytonah, likely to miss that, thanks14:26
creihtyou would find out pretty quick :)14:26
notmynameyou wouldn't miss it for long :-)14:26
*** chmouelb has joined #openstack14:28
*** lmcdowell has joined #openstack14:34
*** lmcdowell has joined #openstack14:34
*** cloudmeat has quit IRC14:34
*** krish has quit IRC14:36
*** annegentle has joined #openstack14:37
sorenchmouel: Apparantly, sudo gets very, very upset with you, if you sudoers file (or an included sudoers file) isn't mode 440.14:46
*** gundlach has joined #openstack14:52
*** cloudmeat has joined #openstack14:52
*** dendrobates is now known as dendro-afk14:53
*** pharkmillups has joined #openstack14:55
*** kashyapc has joined #openstack14:55
uvirtbotNew bug: #644420 in swift "Typo in Getting Started in Swift doc page "developemnt"" [Undecided,New] https://launchpad.net/bugs/64442014:56
*** dendro-afk is now known as dendrobates14:56
jaypipessoren: yep, that error forced me to  boot in recovery mode, into a root shell simply to change the perms, as the nova_sudoers file caused sudo to segfault.  Very annoying.14:58
*** sirp1 has joined #openstack15:01
*** kevnfx has joined #openstack15:03
*** matclayton has quit IRC15:04
*** Podilarius has quit IRC15:07
*** ctennis has quit IRC15:11
jaypipesdendrobates: pls set priority: https://blueprints.launchpad.net/nova/+spec/austin-nosql-datastore-adapter15:11
*** kevnfx has quit IRC15:13
chmouelsoren: yeah prob the permissions in the package should maybe get fixed, have you seen that prb before?15:14
*** Podilarius has joined #openstack15:14
*** jeromatron has left #openstack15:15
jaypipesdendrobates: cheers :)15:18
*** ibarrera has quit IRC15:23
*** tobym has quit IRC15:28
*** gasbakid has joined #openstack15:30
*** calavera has quit IRC15:36
*** pvo has joined #openstack15:49
*** ChanServ sets mode: +v pvo15:49
gundlachcerberus: are you aware of the text conflicts in the rs_auth mergeprop?15:53
gundlach_cerberus_: ^^15:53
*** kevnfx has joined #openstack15:53
_cerberus_gundlach: I am not. I was enroute to the office15:53
_cerberus_I'll check now15:53
gundlachcool,  if you look at the mergeprop page you'll see at the top the list of text conflicts15:54
_cerberus_gundlach: merge conflicts fixed and repushed16:05
gundlachgreat, i'll take a look16:05
jaypipes_cerberus_: FYI, if you set the merge prop to Work in Progress while fixing up merge conflicts, then set it back to Needs Review, all reviewers will be notified to review again.  Just a quick tip ;)16:06
_cerberus_jaypipes: yeah, I did that earlier. I didn't want to start spamming ;-)16:07
jaypipes_cerberus_: :P)16:07
jaypipes_cerberus_: gah, my typing sux today.16:07
*** maplebed has joined #openstack16:16
*** joearnold has joined #openstack16:32
*** metoikos has quit IRC16:33
*** lmcdowell has quit IRC16:43
*** jbryce has joined #openstack16:44
sorenchmouel: I've pushed a fix to the packaging branch. It should be fixed by tomorrow.16:45
*** kevnfx has quit IRC16:48
gundlachvishy: I'm seeing errors in the quota unittests in one of my branches.  One is test_too_many_addresses, which has a note from you saying that this test isn't working properly... should this test be excluded or something?16:48
*** jbryce has quit IRC16:50
gundlach(I'm getting a QuotaError)16:50
*** rlucio has joined #openstack16:56
*** matclayton has joined #openstack16:56
*** zheng_li has quit IRC17:11
*** DubLo7 has joined #openstack17:12
*** chmouelb has quit IRC17:18
*** annegentle has quit IRC17:21
*** dendrobates is now known as dendro-afk17:25
*** hazmat has joined #openstack17:25
*** dendro-afk is now known as dendrobates17:29
*** pharkmillups has quit IRC17:29
*** joearnold has quit IRC17:32
*** matclayton has left #openstack17:43
*** zheng_li has joined #openstack17:50
*** joearnold has joined #openstack17:51
gundlachvishy: nm, i found it.  you were using assertFailure incorrectly, i think -- you were supposed to return its value.  my branch is off twisted so i replaced it with assertRaises.17:55
*** gasbakid has quit IRC17:55
*** pietro has quit IRC17:58
*** joearnold has quit IRC17:58
*** dendrobates is now known as dendro-afk17:58
*** joearnold has joined #openstack17:59
*** kashyapc has quit IRC18:03
notmynamewhat's the bzr equivalent of git stash?18:15
hazmatnotmyname, bzr shelve ?18:15
notmynameand git stash apply == bzr shelve apply?18:16
hazmatnotmyname, bzr unshelve18:16
hazmatnotmyname, bzr shelve --list to see what's shelved.. you can apply a  symbolic name when you shelve18:16
notmynamewhat about git stash clear?18:16
hazmatnotmyname, i'm not a git user so.. dunno.. there isn't a way to kill a shelve content, but you can continually append additional shelves.. and unshelve specific things18:18
hazmatnotmyname, actually bzr shelve --destroy can destroy it..18:19
hazmatbzr shelve --help has the details18:19
notmynameI've got uncommitted changed that conflict with a branch I want to merge in (but I don't want to loose my uncommitted changes) in git, I would git stash; git merge; git stash apply18:19
*** brd_from_italy has quit IRC18:20
hazmatnotmyname, same thing with bzr .. bzr shelve --all -m "premergestuff" && bzr merge ../trunk && bzr unshelve premergestuff18:20
hazmatnotmyname, np18:21
hazmatnotmyname, there's also a very helpful #bzr channel on freenode18:21
*** pharkmillups has joined #openstack18:21
notmynamebut why would I use that when you're here? ;-)18:21
notmynamegood to know. I'll use that in the future18:22
*** joearnold has quit IRC18:25
*** metoikos has joined #openstack18:28
*** Pentheus has joined #openstack18:32
vishygundlach: returning the value from assertFailure solves the issue?18:39
*** annegentle has joined #openstack18:40
gundlachvishy: twisted.trial's documentation mentions that you must return the deferred from your testcase, so i imagine that was the cause of the problem you mentioned in your note.  my branch moves onto eventlet, so i can just use the standrad unittest.TestCase.assertRaises.18:40
vishygotcha, i'll try with returns to see if that fixes it18:42
vishyeveryone: bpython18:42
vishycoolest thing since sliced bread18:42
*** pietro__ has joined #openstack18:42
*** blamar has joined #openstack18:44
*** rlucio_ has joined #openstack18:44
*** rlucio has quit IRC18:46
*** rlucio_ is now known as rlucio18:46
*** skippyish has left #openstack18:48
gundlachwow vishy, thanks; will switch from ipython18:48
gundlachi love the autocomplete ui18:48
*** DubLo7 has quit IRC18:48
*** pietro__ has quit IRC18:50
*** dendro-afk is now known as dendrobates18:51
*** DubLo7 has joined #openstack18:51
gundlachsoren: i'm stuck trying to merge https://code.launchpad.net/~gundlach/nova/controllers-in-api/+merge/34795 again, same error as yesterday: "deleting parent in nova/endpoint".  could you take a look, or point me to someone else to ask?18:51
gundlachi'm getting to the point where i'm spending a couple hours a day just trying to get this to merge, which is silly :)18:51
pvois soren still around?18:52
gundlachdunno, he's still in the room, but idle18:52
pvomtaylor might be lurking ... :)18:52
gundlachah, thx, last i checked mtaylor wasn't in here :)  mtaylor: ^^ ?18:52
mtaylorpvo: aroo?18:52
* gundlach cheers18:52
gundlachmtaylor, i'm in need of some serious hudson-launchpad-fu18:53
mtaylorgundlach: I will do my best to provide such a thing :)18:53
gundlachthx! :)  so i've got two concurrent problems:18:53
vishygundlach: yeah it really rocks18:54
gundlach1) when i try to merge lp:~gundlach/nova/controllers-in-api to lp:nova, sometimes i get the "deleting parent in nova/endpoint" error mentioned above.  i don't know how to convince trunk that it's *ok* that i'm deleting the nova/endpoint/ directory.18:54
*** allsystemsarego has quit IRC18:54
gundlach2) when i do manage to merge, i'm getting unittest errors that look like eventlet 0.9.12 hasn't been installed on hudson yet to replace 0.9.10.  on my local venv, i don't get those errors.18:54
vishygundlach: i just created a branch adding support to nova-manage for bpython18:54
gundlachvishy: :D18:54
gundlachmtaylor: so at the moment i'm trying to merge a branch that intentionally has a unittest fail by printing out eventlet.__version__, to prove to myself that 0.9.12 is really installed :) but #1 above is blocking that experiment at the moment.18:55
mtaylorgundlach: hrm. ok. well, I thought soren was working on getting the eventlet upgrade done... lemme check18:56
mtaylorand the other thing is weird... but lemme look at that too :)18:56
gundlachyep, he's reported that it completed, which is why i'm confused18:56
gundlachas usual, probably my lack of bzr/lp knowledge causing the prob, but i'm stuck hard enough to ask for help :)18:57
mtaylorok. no - eventlet 0.9.10 is still installed18:59
* gundlach steps back from the abyss19:00
mtaylorgundlach: ok. 0.9.12 is now installed19:00
mtaylorgundlach: ok - trying it again.19:06
mtaylorgundlach: oh - didn't you say you pushed a deliberate test case fail?19:07
gundlachyeah, and i just pushed the revert of that.  it looks like my Approve got in before the revert.19:07
gundlachi'll approve again :)19:07
mtaylorI just did19:08
gundlachso it looks like you fixed whatever the "delete parent endpoint" error was about -- however you did it, thank you.19:08
mtaylorany time!19:12
gundlachIT WORKED!19:12
gundlachmtaylor: let me buy you a beer at the summit19:13
creihtannegentle: welcome to your first merge :)19:13
* annegentle makes friends with bzr19:14
mtaylorgundlach: I will allow you to buy me beer :)19:14
vishyannegentle: bzr is a fickle friend19:18
*** ChanServ sets mode: +v _cerberus_19:20
*** mtaylor has quit IRC19:20
*** gustavomzw has quit IRC19:33
*** _anm has joined #openstack19:39
*** anm_ has quit IRC19:41
*** _anm is now known as anm_19:41
*** metoikos has quit IRC19:42
*** krish has joined #openstack19:50
*** brd_from_italy has joined #openstack19:50
dendrobatesRelease meeting in 1 hour in #openstack-meeting20:00
*** jsgotangco has joined #openstack20:13
*** skippyish has joined #openstack20:16
*** pvo_ has joined #openstack20:21
*** joearnold has joined #openstack20:22
_cerberus_So it appears my merge failed. https://code.launchpad.net/~cerberus/nova/rs_auth/+merge/35727 I attempted to add a new model to the db for the API auth tokens. It would seem I need to do something else to facilitate that change20:25
_cerberus_Someone care to point me in the right direction?20:25
*** skippyish has quit IRC20:27
*** skippyish has joined #openstack20:31
gundlach_cerberus_: did these tests pass in your branch, before you tried merging?20:32
_cerberus_gundlach: I failed to run the full suite :-/20:33
_cerberus_My own tests passed, and I forgot to check the outer suite :-P20:34
gundlachand, i've found it useful to "bzr branch lp:nova trunk" then cd to trunk/ and "bzr merge lp:<the branch i'm trying to merge in hudson>" to check if i get the same test failures as hudson does20:34
gundlachah!  well, there you go :)20:34
*** skippyish has quit IRC20:34
*** skippyish has joined #openstack20:35
*** DaFrog has joined #openstack20:35
*** ianw has joined #openstack20:35
*** DaFrog has left #openstack20:35
_cerberus_gundlach: hmmm, I'm likely doing something wrong, but they do all fail in trunk as well20:35
gundlachyour test failures look like something is really basically wrong e.g.20:36
gundlach[ERROR]: nova.tests.access_unittest.AccessTestCase.test_002_allow_none20:36
gundlachTraceback (most recent call last):20:36
gundlach  File "/var/lib/hudson/src/nova/hudson/nova/test.py", line 217, in run20:36
gundlach    self.setUp()20:36
gundlach  File "/var/lib/hudson/src/nova/hudson/nova/tests/access_unittest.py", line 76, in setUp20:36
gundlach    self.context.project = self.project20:36
gundlachexceptions.AttributeError: 'AccessTestCase' object has no attribute 'project'20:36
gundlachdoesn't self.project get set in setUp()?20:36
gundlachmaybe you somehow made setUp() not get run on the tests?20:36
_cerberus_*that* seems really unlikely20:37
gundlachwhen i have test errors in run_tests.sh, i modify run_tests.py to remove all but one "from some_tests import *" and add an "import logging" which seems to be required and then run them again20:37
*** pvo_ has quit IRC20:37
gundlachto help narrow it down.  try that out and see what yo uget.20:38
*** rlucio has quit IRC20:38
*** skippyish has quit IRC20:40
*** skippyish has joined #openstack20:41
*** krzycoder has joined #openstack20:43
*** scottie has quit IRC20:45
edaygundlach: it looks like the cloudpipe API parts are still dependent on tornado/nova.endpoint modules... did you have a branch fixing this yet?20:46
uvirtbotNew bug: #644704 in swift "Change object-updater to track what containers have been updated" [High,New] https://launchpad.net/bugs/64470420:46
gundlachnope, i just got unit tests to pass.  i hadn't realized those were still tied in.20:46
*** scottie has joined #openstack20:46
gundlachi'm not familiar with cloudpipe -- is that something that's going into Austin?20:46
edaygundlach: well, it's been in there for a while, it's the VPN functions around servers20:47
edaygundlach: things like bin/nova-manage still depend on endpoint as well :)20:47
gundlachhmm, i guess they should have had unit tests ;)20:47
gundlachi'll take a look -- i'm working on redoing access_unittest now that rbac has gone away20:48
edaygundlach: yeah, probably should have. either way, should probably get this fixed up since trunk is broken, I'm going to poke at it some20:48
gundlachthat would be great!20:48
edaycan remove these tornado deps as well20:49
gundlachif you make a branch lemme know and i'll review it asap20:49
edayvishy: I assume you guys still use/need cloudpipe?20:51
*** ohiocitrix has joined #openstack20:52
ohiocitrixIs there an openstack meeting coming up in a few or did I miss it?20:53
creihtohiocitrix: in 6 minutes20:54
creihtin #openstack-meeting20:54
ohiocitrixOK, will go there. appreciate it20:54
*** ohiocitrix is now known as spectorclan20:55
*** adjohn has joined #openstack20:56
dendrobatesRelease meeting in 4 min in #openstack-meeting20:56
*** sparkycollier has joined #openstack21:01
*** JimCurry has joined #openstack21:01
*** lmcdowell has joined #openstack21:02
*** ctennis has joined #openstack21:03
*** ctennis has joined #openstack21:03
edaygundlach: it looks like in the port to wsgi, the ec2 URL maps dropped the cloudpipe controllers all together:21:05
eday            (r'/cloudpipe/(.*)', nova.cloudpipe.api.CloudPipeRequestHandler),21:05
eday            (r'/cloudpipe', nova.cloudpipe.api.CloudPipeRequestHandler),21:05
gundlachah yeah, i added a TODO which may have gotten lost in the 70 merge attempts which said "I'm not adding cloudpipe until i hear that we absolutely need it"21:06
*** jdarcy has quit IRC21:06
edayhmm, ok. probably should have confirmed before it hit trunk :)21:06
gundlachyeah, i hadn't realized how much of the code wasn't covered by unittests.21:07
gundlachpardon me for leaving so much stuff messy in this merge, apparently -- the merge has been a real headache :)21:07
edayyeah.. testing def needs to be improved21:07
zulsoren: i noticed in the trunk you need a newer python-eventlet are you going to ask for a FFE?21:11
sorenzul: I expect to, yes.21:12
zulsoren: k thanks21:12
jaypipesvishy: if time, #openstack-meeting.  would be good to get status updates from anso :)21:15
zulsoren: if you need help let me know21:15
sorenzul: Will do, thanks.21:16
*** scottie has quit IRC21:17
*** cloud0 has quit IRC21:20
*** rlucio has joined #openstack21:22
*** kevnfx has joined #openstack21:22
*** scottie has joined #openstack21:28
*** jkakar has quit IRC21:31
*** annegentle has quit IRC21:32
*** jkakar has joined #openstack21:33
*** gustavomzw has joined #openstack21:41
vishyeday: we def. need cloudpipe but we are still using cloud.py at the moment21:41
vishygundlach: cloudpipe endpoint could be its own app21:42
gundlachvishy: fine with me; i don't need to get cloudpipe off of tornado, i just needed to get the API off of tornado.  if cloudpipe doesn't expose an API, we're set.21:43
gundlach(public API)21:43
gundlachvishy: re cloud.py -- do you mean api.ec2.cloud.py?  Be aware that's off of Twisted now21:43
edayvishy, gundlach: It looks like we'll probably want to do api/cloudpipe, and then create a generic run_instances function somwhere that both ec2/rs/cloudpipe can call21:43
gundlacheday: remind me -- run_instances fires up some new virtual servers?21:44
edaygundlach: yup21:44
edaygundlach: cloudpipe currently takes a controller and uses that to start some, but we should have it using a lower-level function most likely21:45
edaygundlach: we'll want to create one anyways for ec2/rackspace APIs to share21:45
gundlachif i understand the arch correctly (and I don't ;), it seems to make sense that all 3 of ec2/rs/cloudpipe would put the same message onto the queue to start the process of spawning instances.  wouldn't that be the common API they share rather than some class/function?21:45
*** spectorclan has quit IRC21:46
*** scottie has quit IRC21:46
gundlachyou know what, don't even bother answering my question -- i'm sure you understand better what's going on, as i have barely looked at servers impl :)21:46
edaygundlach: yes, right now cloudpipe calls nova.endpoint.cloud.run_instances (old API call that doesn't exist now). We'll want both of those and the rackspace "start servers" call all using the same method/function/whatever to send the same message to the scheduler21:47
*** jsgotangco has quit IRC21:48
gundlachendpoint.cloud.run_instances == api.ec2.cloud.run_instances; maybe it's sufficient to point RS API to that function, then21:48
gundlach(and to point cloudpipe there as well)21:48
gundlachand, now that i'm seeing that cloud.py may be used for more than just EC2 api, maybe api.ec2.cloud should be moved to api.cloud or something21:49
edaygundlach: yup, pretty much. although I don't think the canonoical function should live in api.ec2, probably move to nova.compute somewhere21:49
*** brd_from_italy has quit IRC21:50
edaygundlach: and all api/*/cloud can call nova.compute.api.run_instances() or something21:50
gundlachsounds good to me21:50
edaybut I'm working on a branch that fixes cloudpipe and removes tornado for good21:51
vishyso lets say that i accidentally did a bzr pull in my parent directory21:52
vishyhow do i fix it?21:52
vishyjust completely hypothetically speaking21:52
*** adjohn has quit IRC21:52
vishyand for the record it is really really dumb that bzr lets me do that21:53
gundlachvishy: about as dumb that it lets you type 'bzr revert' in the parent directory?21:53
pvohypothetically... :)21:53
vishyit does21:53
vishyand it doesn't do anything useful21:53
vishyit doesn't actually revert in the parent dir21:54
gundlachparent dir meaning nova/, inside of which live your branches and trunk/ ?21:54
vishyso no ideas?21:54
dendrobatesperhaps a question for #bzr21:54
vishysigh ok i guess i'll clean it up by hand grumble21:55
*** gondoi has quit IRC21:57
*** hazmat has quit IRC22:00
*** Xenith has quit IRC22:03
*** sparkycollier has quit IRC22:03
*** Xenith has joined #openstack22:03
*** sparkycollier has joined #openstack22:04
*** gundlach has quit IRC22:04
vishyso user error22:06
vishyi used bzr init instead of init-repo22:06
*** joearnol_ has joined #openstack22:11
*** joearnold has quit IRC22:15
*** pvo has quit IRC22:22
*** littleidea has joined #openstack22:23
*** kevnfx has quit IRC22:28
*** mtaylor has joined #openstack22:29
*** ChanServ sets mode: +v mtaylor22:29
*** JimCurry has left #openstack22:33
*** rnirmal has quit IRC22:33
*** littleidea has quit IRC22:35
*** Ryan_Lane has joined #openstack22:37
*** abecc has quit IRC22:37
*** sparkycollier_ has joined #openstack22:37
*** ianw has quit IRC22:38
*** sparkycollier has quit IRC22:40
*** sparkycollier_ is now known as sparkycollier22:40
*** scottie has joined #openstack22:41
*** pharkmillups has quit IRC22:57
*** anm_ has quit IRC23:10
*** joearnol_ has quit IRC23:14
*** joearnold has joined #openstack23:14
*** Cybodog has quit IRC23:16
*** klord has quit IRC23:22
*** miclorb_ has joined #openstack23:27
*** cloudmeat has quit IRC23:27
vishygundlach: are you here? I'm getting test errors in S3APITestCase about twisted being activated multiple times23:28
vishyand reactor being unclean23:28
vishyi need eventlet 9.12?23:29
*** sophiap has quit IRC23:32
*** sophiap has joined #openstack23:36
rluciois anyone planning on updating the debian-packaging branch anytime in the near future?  The branch is already pretty far out of date23:39
*** joearnol_ has joined #openstack23:44
*** joearnold has quit IRC23:47
*** johnbergoon has joined #openstack23:47
vishyrlucio: yeah, we need to do that23:55
vishyrlucio: i was hoping to convince soren or monty to do it, but I might try and tackle it if they aren't available23:56
*** gustavomzw has quit IRC23:56

Generated by irclog2html.py 2.14.0 by Marius Gedminas - find it at mg.pov.lt!