Thursday, 2011-01-06

*** dubsquared has quit IRC00:06
*** kevnfx has quit IRC00:08
*** dfg_ has quit IRC00:09
*** nelson__ has quit IRC00:17
*** nelson__ has joined #openstack00:18
*** rogue780 has joined #openstack00:19
*** normanrichards has quit IRC00:20
Ryan_Lanefor a user to be able to create instances only in a specific project, what roles do they need to be in?00:23
*** hggdh has quit IRC00:30
*** hggdh has joined #openstack00:30
*** kevnfx has joined #openstack00:33
*** kashyapc has quit IRC00:34
*** littleidea has quit IRC00:35
*** littleidea has joined #openstack00:35
vishyRyan_Lane: projectadmin developer or sysadmin00:36
vishyfor dev and sys they need the project role and the global role00:36
vishyany upstart geniuses around?00:36
Ryan_Lanevishy: as global roles, or project roles?00:38
Ryan_Laneit doesn't even seem to be searching for project roles00:38
Ryan_Lanealso. I'm just noticing this...00:38
Ryan_Laneis there any way for me to give an instance a name on creation?00:38
*** rogue780 has quit IRC00:39
Ryan_LaneI was using DisplayName, but now nova seems to assign something to that00:39
*** jc_smith has quit IRC00:39
*** matiu_ has joined #openstack00:40
Ryan_Lanenow I'm getting names like "Server 1235955006" which is completely worthless00:40
*** rogue780 has joined #openstack00:40
*** winston-d has left #openstack00:41
*** winston-d has joined #openstack00:41
*** matiu has quit IRC00:43
*** matiu_ is now known as matiu00:43
*** normanrichards has joined #openstack00:43
winston-dcreiht: hi~00:43
*** normanrichards has quit IRC00:44
notmynamewinston-d: what's up?00:47
winston-dnotmyname: I'm still trying to understand low PUT performance of my swift setup.00:47
notmynamehave you seen any errors in you logS?00:48
winston-dwell, not obvious ones.00:48
notmynameyou asked about separating the proxy logs from syslog yesterday. did you get that working?00:49
winston-dand when I try to follow the instructions here to log proxy server, i found RHEL 6 is using rsyslog instead of syslog-ng00:49
*** kashyapc has joined #openstack00:50
notmynameok. you should still be able to see your proxy logs by grep'ing syslog00:51
Ryan_Lanevishy: yeah. get_user_roles is never called...00:51
Ryan_Laneunless I'm debugging poorly :)00:51
nelson__argh: swift-auth-add-user -K leakywiks -a system root testpass00:52
notmynameI've only got about 10 more minutes until I have to go for the evening. what have you tried? do you suspect any part of your setup to be the main bottleneck?00:52
nelson__Update failed: 503 Service Unavailable00:52
Ryan_Laneit looks like the only roles ever looked at are global roles00:52
notmynamenelson__: auth.db permissions?00:53
winston-dnotmyname: my working day just began. :)  let me do more analysis and leave messages here.00:53
*** kevnfx has quit IRC00:53
notmynamewinston-d: ok. what TZ are you in?00:53
vishyRyan_Lane: checking00:53
winston-di'm in GMT+800:54
nelson__notmyname: geezums, dont' you think that should be in the do.... oh. oops.00:54
notmynamenelson__: heh. it's fixed in trunk, but swift 1.1 still has the bug I think00:54
nelson__well yeah, seemingly. :)00:54
notmynamewinston-d: makes it hard to talk to devs almost exactly on the other side of the planet from you (GMT-6)00:55
vishyRyan_Lane: has_role in auth/manager.py00:55
nelson__hrm. that doesn't fix it.00:55
Ryan_Lanevishy: that checks global groups only00:55
notmynamenelson__: it should have logged something00:55
vishyno the last call checks project role00:56
vishyit shortcuts if the global role isn't set00:56
winston-dnotmyname: I know.  I complained that before.  :)  using a-synced way such as mailing list should be easier but clearly it seems to me openstack mailing list is not a good place00:56
nelson__notmyname: okay, that's weird. the second time I ran it, it worked.00:56
vishyRyan_Lane: it checks global role first, then checks project role00:56
Ryan_LaneI don't see how...00:57
Ryan_Lanethe last call is "drv.has_role"00:57
nelson__I'm not gonna worry about it. I have my auth key now.00:57
notmynamewinston-d: we all get the mailing list. it is more difficult for live break fixing, but feel free to use it if it's more convenient00:57
vishyRyan_Lane: the docstring is wrong, it should say the "intersection"00:57
Ryan_LaneI'm not even looking at the documentation :)00:57
notmynamenelson__: you should only have to set it one time (per swift reset)00:57
Ryan_Lanevishy: has_role in the ldap driver only checks global roles00:57
vishywho did that?00:58
notmynameok. gotta go. time to put the kids to bed00:58
Ryan_Laneno idea00:58
Ryan_Lane(I hope it wasn't me)00:58
winston-dnotmyname: OK.  I'll try both ways.00:58
vishyRyan_Lane no00:58
vishyrole_to_dn switches on project_id00:59
vishyif project_id is set it looks at the role under project00:59
Ryan_Laneah. I wonder if I screwed something up in my new branch00:59
vishyRyan_Lane: yeah probably you aren't respecting project_id in __role_to_dn01:00
Ryan_Laneah. very very likely01:01
Ryan_Lanehah. I am indeed01:02
Ryan_Lanewait... no I think I'm doing it right after all :(01:02
*** normanrichards has joined #openstack01:03
winston-dis there anyway i can suppress the log generated by DNSMASQ?01:03
vishyRyan_Lane: it looks right to me01:04
Ryan_Lane__role_to_dn gets the project dn via __project_to_dn, and appends cn=project_id to it01:04
Ryan_Lanevishy: you looking at my newest branch?01:04
vishyRyan_Lane: looking at the bug branch01:04
vishyRyan_Lane: are you seeing an actual problem?01:05
Ryan_LaneI'm trying to create instances01:05
Ryan_Laneand getting a 40101:05
Ryan_Lanevia the ec2 api01:05
vishyusing eucatools?01:05
Ryan_LaneI'm also looking at the ldap logs, which is why I was asking about this :)01:05
vishyis the user a member of project?01:06
Ryan_LaneIn the logs I don't see it looking for roles under the projects01:06
vishywell does the user have the global role?01:06
Ryan_LaneI manually added a role under the project01:06
Ryan_Laneit works if I add the user to a global role01:06
Ryan_Lanelike cloudadmin01:06
vishyit will only check for project role if it also has the global role01:06
vishyneeds global developer and project developer01:07
Ryan_Lanedoesn't that kind of defeat the purpose?01:07
vishythis is supporting a specific use case01:07
vishywhich seems to confuse everyone01:07
vishyI think i explained it to you at one point01:07
Ryan_LaneI'm obviously still confused about it :D01:08
vishymight make it clearer if the docstring in has_role said INTERSECTION in big letters instead of union01:08
Ryan_Laneis cloudadmins a special group?01:08
vishyand explained that you need both the user level (global) role and the project role01:08
vishycloudadmin can do everything01:08
Ryan_Laneeven if the project role doesn't exist?01:09
vishyitsec is special too01:09
Ryan_Laneok. my confusion is solved then :)01:09
vishythere is a flag for global_roles01:09
vishywhich includes cloudadmin and itsec01:09
vishy"global_roles" don't need matching project roles01:09
*** Jordandev has joined #openstack01:13
Ryan_Lanevishy: thanks for solving my problem (which was my confusion)01:16
vishyRyan_Lane: I'm thinking that we should switch the default01:16
Ryan_Lanewhat do you mean?01:16
vishyRyan_Lane: just make has_role return the specific global or project role01:17
vishyRyan_Lane: and by default check for either01:17
Ryan_Laneso that it isn't an intersection?01:17
vishyso that we're checking for the union, that is the way most people expect it to work01:18
*** Jordandev has quit IRC01:18
Ryan_Lanethat makes more sense01:18
Ryan_Lanewas this a nasa or rackspace specific case?01:18
vishywe can specify a flag to use intersection if desired01:18
vishynasa specific01:18
* Ryan_Lane nods01:18
Ryan_Laneit would make my coding way easier :)01:19
vishymaybe make a bug: role checking should use union by default instead of intersection01:19
vishythen one of us can knock it out soon01:19
Ryan_Laneon it01:19
Ryan_Lanecool. thanks01:19
*** joearnold has quit IRC01:20
*** aliguori has quit IRC01:22
Ryan_Lanethat'll solve one of my problems01:23
Ryan_Lanenow I need to figure out how to assign a DisplayName to the instances on creation. that somehow changed recently01:24
*** aliguori has joined #openstack01:24
dendrobates vishy: I'll check back in an hour or so for that ppa code drop.01:27
*** ArdRigh has joined #openstack01:28
*** Jordandev has joined #openstack01:30
uvirtbotNew bug: #697936 in nova "role checking should use union by default instead of intersection" [Undecided,New]
JordanRinkeany openstack admins here that can give me a quick add to the planet openstack team?01:34
* dirakx managed to get a zipfile of nova ;). 01:36
dendrobatesJordanRinke: I can01:38
dendrobatesJordanRinke: done01:40
*** dendrobates is now known as dendro-afk01:42
openstackhudsonProject nova build #361: SUCCESS in 1 min 15 sec:
JordanRinkedendro-afk: much appreciated01:44
Ryan_Laneugh. assigning of displayName is broken in the ec2 api :(01:46
*** rogue780 has quit IRC01:47
*** dendro-afk is now known as dendrobates01:48
*** maplebed has quit IRC01:49
*** blpiatt has quit IRC02:00
*** anotherjesse has quit IRC02:01
*** littleidea has quit IRC02:04
*** JohnFury has joined #openstack02:06
JohnFurygod evening people and happy new year to people all around the world02:09
JohnFurygood evening i say...02:10
*** littleidea has joined #openstack02:10
*** anotherjesse has joined #openstack02:16
*** jdurgin has quit IRC02:17
*** dendrobates is now known as dendro-afk02:24
*** dendro-afk is now known as dendrobates02:25
*** JohnFury has quit IRC02:28
sandywalshhey guys, a particular operation I'm running through virt/xenapi_conn::wait_for_task is throwing an exception (as expected). My callback gets the notification, but it appears LoopingCall is attempting the operation again.02:29
sandywalshI get this error:
*** schisamo has quit IRC02:30
sandywalshany idea how to prevent this?02:30
*** hadrian has quit IRC02:31
sandywalshgot it ... ignore :)02:32
*** bvierra has joined #openstack02:34
*** blpiatt has joined #openstack02:34
bvierrahey all looking to use openstack to replace XenServer02:34
bvierrais there anywhere that has a list of hardware requirements / recommendations?02:35
sandywalshbvierra, openstack works with xenserver, but doesn't replace it. It controls it.02:35
anotherjesseXenServer is a hypervisor - we support it - along with KVM02:35
alekibangoprobably no, that should be in deployment manual02:35
alekibangobut memory = good02:36
bvierrahmm ok02:36
alekibangohaving 128 MB memory is not enough :)02:36
* bvierra goes back to reading02:36
alekibangoi would say at least 32GB is good idea02:36
alekibangobut, nova can work with different hw...02:37
alekibangoit all depends on what you need it for02:37
anotherjessealekibango / bvierra yeah, it depends on use cases02:42
anotherjessecompute heavy jobs vs data heavy jobs vs web vs "enterprise" ...02:42
alekibangoanotherjesse: still, some examples of hw for different common cases should be documented02:42
bvierraanotherjesse, so you still need to use some hypervisor along with it, do you have a recommendation other than XenServer?02:42
alekibangobvierra: kvm is supported02:43
anotherjessebvierra: KVM works nice ... my only problem with it is that when you attach a volume, users can't accurately specify /dev/vd* entry it shows up on :(02:43
creihtwinston-d: howdy02:51
Ryan_LaneI'm tracking down a displayname issue, and not successfully nailing it down yet02:51
winston-dcreiht: :)02:51
creihtahh just saw your pm02:52
*** daleolds has quit IRC02:52
Ryan_LaneI believe it's in the db instance creation... compute/ -> create_instances gets the display_name properly, but when it creates it in the database, and checks for display_name in the result, it doesn't find it02:53
Ryan_Laneso it assigns a default display_name02:53
*** freeplant has joined #openstack03:10
*** cyonyx has joined #openstack03:10
*** lamboric has joined #openstack03:13
*** perestrelka has quit IRC03:17
*** trin_cz has quit IRC03:41
*** dirakx has quit IRC03:52
* Ryan_Lane sighs03:53
Ryan_LaneI upgraded my trunk packages, and for some reason no service can talk to the database03:53
*** ArdRigh has quit IRC04:15
*** freeplant has quit IRC04:17
*** lamboric has quit IRC04:28
*** dragondm has joined #openstack04:39
*** kashyapc has quit IRC04:39
*** jbryce_ has joined #openstack04:53
*** dragondm has quit IRC04:55
*** jbryce has quit IRC04:56
*** jbryce_ is now known as jbryce04:56
*** kashyapc has joined #openstack05:00
*** kevnfx has joined #openstack05:04
*** ArdRigh has joined #openstack05:04
*** littleidea has quit IRC05:05
*** damon__ has quit IRC05:11
*** jimbaker has quit IRC05:16
*** ArdRigh has quit IRC05:34
*** kevnfx has quit IRC05:36
*** littleidea has joined #openstack05:38
*** f4m8_ is now known as f4m805:47
*** kevnfx has joined #openstack05:53
*** kevnfx has quit IRC06:04
*** lunablue has joined #openstack06:07
*** miclorb has quit IRC06:08
*** dendrobates is now known as dendro-afk06:15
*** lunablue has quit IRC06:16
*** ramkrsna has joined #openstack06:29
*** JohnFury__ has joined #openstack06:37
*** ArdRigh has joined #openstack06:37
JohnFury__ttx or anyone: do you know anyone who can help regarding creating a CentOS image compatible with openstack?06:38
*** blpiatt has quit IRC06:44
*** samhamilton has joined #openstack07:01
*** samhamilton has quit IRC07:13
*** jq889 has joined #openstack07:15
*** JohnFury__ has quit IRC07:20
*** Jordandev has quit IRC07:25
ttxvishy: about the Ubuntu packaging branch, I can commit there as well, like any other Ubuntu core dev07:34
ttx(or MOTU)07:34
ttxvishy: feel free to propose branches and ping me for merging07:35
ttx(though I was in bed at the same time as soren :)07:35
ttxAbout pep8: the rule is there to ensure style consistency... If you look at Eucalyptus code you would see how much bad style consistency can hurt code readability07:36
*** BK_man has joined #openstack07:41
*** jq889 has quit IRC07:53
*** allsystemsarego has joined #openstack07:54
*** leted has joined #openstack07:54
*** miclorb_ has joined #openstack08:08
*** czajkowski has quit IRC08:16
*** czajkowski has joined #openstack08:16
*** leted has quit IRC08:18
*** BK_man has quit IRC08:29
*** BK_man has joined #openstack08:34
*** elasticdog has quit IRC08:35
*** DeAnza has quit IRC08:38
*** skrusty has quit IRC08:43
*** elasticdog has joined #openstack08:44
*** littleidea has quit IRC08:45
*** MarkAtwood has quit IRC08:48
*** DigitalFlux has joined #openstack08:55
sorenttx: nova-core people can commit there, too, by the way.08:55
*** BK_man has quit IRC08:55
sorenttx: Isn't Eucalyptus actually typical "enterprise" java style code?08:56
*** littleidea has joined #openstack08:56
ttxsoren: ah? sounds pretty open :)08:56
sorenttx: Well, things don't automatically trickle into Ubuntu just because it's in that branch.08:56
ttxsoren: I saw the Enterprise Java code written with coding guidelines08:56
ttxit can look better08:57
*** skrusty has joined #openstack08:57
sorenttx: Considering how much time I've spent working with Eucalyptus, I've spent surprisingly little time looking at the code. Over the last two years, I have have looked at it for a total of 10 minutes. Never felt smarter afterwards.08:57
ttxthough I find pep8 does quite a good job (for a set of rules) to endure consistency08:57
sorenOk, maybe 20 minutes.08:58
ttxsoren: when a patch touches 1000 lines just because people have a different conception of indent, that's a fail in my book :)08:58
sorenBut no more than that. It always felt futile.08:58
ttxsoren: It kinda made me feel smart.08:59
ttxsoren: as in: "even me can write better code than that"08:59
ttxSometimes I feel dumb looking at openstack code :)09:01
*** ramkrsna has quit IRC09:09
*** littleidea has quit IRC09:11
*** czajkowski has quit IRC09:17
*** czajkowski has joined #openstack09:17
*** ramkrsna has joined #openstack09:22
*** perestrelka has joined #openstack10:00
*** trin_cz has joined #openstack10:12
ttxsoren: what's the status for bexar-ec2-soap-api ? will be proposed today, might be proposed later with an exception, or deferred to cactus ?10:22
sorenMight be proposed later with an exception.10:22
ttxok, marking "slow progress" then.10:22
sorenSome of it's done (I can succesfully do a DescribeImages call, for instance).10:23
*** trin_cz has quit IRC10:23
ttxfeel free to push a branch to prove that :)10:23
ttx(or link it if it already exists)10:23
sorenOh, I thought I did.10:23
sorenWill do.10:23
*** lcfseth has joined #openstack10:24
soren..but I can't pass arguments in yet (deserializin that SOAP stuff is no fun at all), nor do I yet have the authentication working (I'm just authenticating all incoming requests as "(soren, soren)") :)10:24
ttx"don't try this at home"10:24
lcfsethhi, may I ask what you're talking about? just logged in :)10:25
ttxlcfseth: see /topic10:25
ttxlcfseth: openstack development in general,  bexar-ec2-soap-api in that specific discussion10:26
sorenttx: Yeah, very much "don't try this at home" :)10:26
*** miclorb_ has quit IRC10:27
*** kashyapc has quit IRC10:30
*** ArdRigh has quit IRC10:30
sorenttx: Pushed and linked.10:31
*** brd_from_italy has joined #openstack10:50
*** cw has quit IRC10:57
*** cw has joined #openstack11:05
*** adiantum has joined #openstack11:06
*** kashyapc has joined #openstack11:08
*** trin_cz has joined #openstack11:14
*** matiu has quit IRC11:19
*** anotherjesse has quit IRC11:23
*** Jordandev has joined #openstack11:40
*** lcfseth has left #openstack11:41
*** Lcfseth has joined #openstack11:46
*** brd_from_italy has quit IRC11:50
*** arthurc has joined #openstack11:51
*** Lcfseth has quit IRC11:52
*** adiantum has quit IRC11:54
*** zul has quit IRC12:01
* soren lunches12:05
*** befreax has joined #openstack12:10
*** Jordandev has quit IRC12:21
*** ramkrsna has quit IRC12:26
*** Jordandev has joined #openstack12:34
*** Lcfseth has joined #openstack12:35
*** ctennis has quit IRC12:38
*** Lcfseth has left #openstack12:43
*** befreax has quit IRC12:52
sorenttx: Juts fyi: If we need new packages built (because someone updated the packaging), clicking the "Build now" thing on the nova-ppa job on Hudson now works.12:52
ttxsoren: ack -- I'll probably need some creds on Hudson to do that, I suspect12:53
sorenttx: Before, it would have generate another 2011.1~bzr521-0ubuntu0ppa1~lucid1, for instance (i.e. not have incremented the lucid1 bit to lucid2).12:53
sorenttx: Oh, you don't have hudson powers?12:53
ttxI don't think so12:54
sorenttx: Have you set up a Hudson account?12:54
sorenttx: If so, I can bless you.12:54
ttxI'll do that right away12:54
* ttx needs a break from looking at Launchpad code interfaces12:54
* soren had similar reasons for doing this in the first place :)12:54
ttxsoren: done, as "ttx"12:56
*** openstackhudson has quit IRC12:57
sorenttx: Great. If you go to now, you should see a "Build now" link on the left.12:58
sorenttx: Possibly in French.12:58
*** openstackhudson has joined #openstack12:58
ttxsoren: yes, I see it.12:58
ttxback to WHY oh WHY won't LP export my linked_branches collection in its *$!#% API12:59
*** ctennis has joined #openstack13:04
*** ctennis has joined #openstack13:04
sorenHyper-V branch approved.13:05
* soren sits back and watches Hudson do its thing.13:05
*** zul has joined #openstack13:09
*** adiantum has joined #openstack13:10
*** irahgel has joined #openstack13:11
*** gasbakid|2 has joined #openstack13:13
openstackhudsonProject nova build #362: SUCCESS in 1 min 22 sec:
*** westmaas has joined #openstack13:16
*** jbryce has quit IRC13:17
*** westmaas has quit IRC13:19
* soren silently shakes his head13:36
sorenttx: Ok, so yesterday, I noticed someone mention Hybridfox on Twitter (in the context of OpenStack).13:36
sorenttx: I wondered what it was, and went to look it up.13:37
sorenttx: Turns out it's a fork of Elasticfox that works with Eucalyptus.13:37
sorenttx: Apparantly, fixing Eucalyptus was too hard.13:37
sorenttx: Incidentally, fixing Nova wasn't. So I did.13:37
sorenSo now Elasticfox pretty much works with Nova. Except for one small hiccup.13:37
sorenttx: Namely the base-36 vs base-16 ID's.13:38
sorenI did notice something interesting while working on this, though.13:41
soren(My stuff is in lp:~soren/nova/elasticfox-compatibility, by the way)13:41
sorenThe image manifest says which kind of virtualisation is used.13:42
sorenUsually, it's paravirt.13:42
*** guigui has joined #openstack13:43
soren...but we can use that do designate an image to be "fully virtual" or "UML" or whatnot.13:43
sorenOr, "fully virtualised with virtio disk"13:43
*** fabiand_ has joined #openstack13:43
sorenttx: Thought I'd mention it, if you were still considering working on the hybrid virt tech stuff.13:44
ttxyes, thanks :)13:44
*** hggdh has quit IRC13:47
*** dendro-afk is now known as dendrobates13:50
*** nelson__ has quit IRC13:51
*** nelson__ has joined #openstack13:51
*** hggdh has joined #openstack13:53
*** blpiatt has joined #openstack13:53
*** westmaas has joined #openstack13:53
sorenttx: You'll have to extend euca-bundle-image to allow it first, though.13:55
*** aliguori_ has joined #openstack13:57
sandywalshOS API merge prop ... anyone care to review? I'm sure there'll be much discussion:
sorensandywalsh: For a start, please set a commit message that will be useful in a changelog.14:01
*** aliguori has quit IRC14:01
sorensandywalsh: Can you elaborate on the krm_mapping.json.sample thing?14:03
sorensandywalsh: I think I understand the motivation, but I'm not sure I understand the approach taken here.14:03
sorensandywalsh: We want to expose what Cloud Servers exposes now, right? That's a numeric ID that maps to a kernel/ramdisk/filesystem tuple.14:04
sorensandywalsh: That sample json seems to map a "machine" key to a kernel/ramdisk 2-tuple.14:05
*** arthurc has quit IRC14:05
sorensandywalsh: Either "machine" refers to the filesystem, in which case I'm not sure how the integer ID is going to be deterministically exposed to clients, or the "machine" thing is the integer ID, in which case I don't understand how the filesystem is chosen.14:06
sandywalshsoren, will update commit message14:06
sorensandywalsh: ta14:07
sandywalshsoren <still reading>14:07
sandywalshsoren, ah. Well OS API deals with flavor (disk size, ram, etc) and OS (linux, etc), while EC2 uses the kernel, machine, ramdisk tuple14:08
sandywalshsoren, while we have a concept of flavor in OS API, rackspace dynamically builds the disk so that's not needed and kernel is part of the image14:09
sandywalshsoren, vs. machine + kernel14:09
sandywalshsoren, so, the mapping is needed (short term). I suspect this will all go away with glance14:10
sandywalshsoren, the other issue is that OS API wants a numeric code for os. EC2 has the ami-abcdef format14:10
sandywalshsoren, so I hash the ami-xxxx to the numeric version14:11
*** arthurc has joined #openstack14:11
uvirtbotNew bug: #698123 in nova "Pausing a Paused instance via OS API causes error" [Undecided,New]
sandywalshsoren, that way, existing clients can still work. Admittedly, the number is unwieldy14:11
*** blpiatt has quit IRC14:11
*** guigui has quit IRC14:11
sandywalshsoren, arguably I could have mapped any of kernel, machine or ramdisk to OS ID.14:12
sandywalshsoren, I just needed some anchor14:12
* sandywalsh over14:13
*** DigitalFlux has joined #openstack14:15
sorensandywalsh: Maybe  I'm missing the purpose of the mapping file. I'll keep reading and get back to you if it's still not clear :)14:16
sandywalshsoren, I'll be here.14:16
sandywalshdubs, I'll merge against trunk again14:17
*** MarkAtwood has joined #openstack14:20
*** littleidea has joined #openstack14:20
*** ppetraki has joined #openstack14:23
*** normanrichards has quit IRC14:32
*** abecc has joined #openstack14:32
*** rogue780 has joined #openstack14:39
*** hadrian has joined #openstack14:40
*** littleidea has quit IRC14:44
sorenYay, maverick packages work again.14:44
sorenWell, they build, at least.14:44
*** Jordandev has quit IRC14:52
*** littleidea has joined #openstack14:56
*** f4m8 is now known as f4m8_14:57
colinnichAny Ubuntu gurus know why can't find python-eventlet or python-devel packages for apt-get install? python-setuptools python-pastedeploy python-webob python-xattr all worked fine. I had a similar problem before with eventlet, but got python-devel to install and could use easy_install. But now I can't find python-devel either. Fresh install of Ubuntu 10.04.1 LTS.15:01
*** rogue780 has quit IRC15:01
*** littleidea has quit IRC15:01
sorencolinnich: There was never a python-devel package in Ubuntu.15:02
colinnichsoren: oh well, maybe that's not how I did it :-)15:02
sorencolinnich: Which version of Ubuntu are you running? We have PPA's that have the relevant dependencies (like eventlet).15:02
sorencolinnich: "sudo add-apt-repository ppa:nova-core/trunk" should do it15:02
*** rogue780 has joined #openstack15:03
sorencolinnich: Assuming you're running lucid, maverick, or natty.15:03
colinnichsoren:  10.04.01. Wasn't using the PPA's as I wanted a custom branch (of swift), but I'll give it a go15:03
sorencolinnich: Oh, sorry, I keep assuming people want Nova :)15:03
dabocolinnich: did you apt-get update first? (sorry if the suggestion is insulting, but just checking)15:04
colinnichsoren: yes, it does seem popular15:04
colinnichdabo: no, I'll try that now15:04
colinnichdabo: didn't help -
*** troytoman has joined #openstack15:07
dabocolinnich: perhaps python-eventlet isn't available in 10.04? I'm on 10.1015:07
notmynamecolinnich: echo "deb lucid main" >/etc/apt/sources.list.d/swift-core-ppa-lucid.list15:10
colinnichnotmyname: that's it, thanks John15:12
colinnich(as always :-))15:12
*** mdomsch has joined #openstack15:14
sorennotmyname: It's better to use "add-apt-repository ppa:swift-core/ppa"15:17
notmynamesoren: good to know15:18
sorennotmyname: It also imports the gpg signing keys, for instance.15:18
notmynameah yes. that's step 2 on
colinnichsoren: yeah I did get warnings about that15:18
*** mdomsch has quit IRC15:19
colinnichsoren: and that way has indeed imported the keys, thanks15:19
*** gondoi has joined #openstack15:20
*** fabiand_ has quit IRC15:21
*** ccustine has joined #openstack15:28
*** hadrian has quit IRC15:28
*** abecc has quit IRC15:38
*** spectorclan has joined #openstack15:41
*** littleidea has joined #openstack15:43
*** WonTu has joined #openstack15:44
*** WonTu has left #openstack15:44
*** kevnfx has joined #openstack15:45
*** dagger has left #openstack15:45
*** hadrian has joined #openstack15:48
*** grizzletooth has joined #openstack15:49
*** kevnfx has quit IRC15:53
*** dragondm has joined #openstack15:54
*** blamar has joined #openstack15:55
*** grizzletooth has left #openstack15:56
*** jimbaker has joined #openstack16:06
*** rogue780 has quit IRC16:06
*** arthurc has quit IRC16:08
ttxsirp-: how is xs-snapshots coming along ?16:08
*** jimbaker has quit IRC16:09
*** jimbaker has joined #openstack16:09
sirp-made good progress, cleaning up the branch right now and planning to propose the merge later today16:12
sirp-ttx: ^^^16:13
ttxsirp-: perfect !16:13
*** dfg_ has joined #openstack16:14
*** zaitcev has joined #openstack16:21
*** troytoman has quit IRC16:22
*** kevnfx has joined #openstack16:25
*** johnpur has joined #openstack16:26
*** ChanServ sets mode: +v johnpur16:26
vishysirp-: I think i can make kvm snapshots pretty quickly on top of my cow branch16:27
vishysirp:- are the upper layers and interfaces  finished already / in trunk?16:28
vishysirp-: ^^^16:28
sirp-vishy: for the most part they are, you would just need to make the appropriate changes in libvirt_conn.py16:29
sirp-vishy: it should just work16:29
vishysirp-: ok that is what i was hoping.  I'll try and get something working real quick16:29
ttxvishy: do you happen to know if anotherjesse worked on aws-api-validation ? I'm about to defer it to Cactus16:30
sirp-vishy: cool, you may end up having to write a little bit of code to put the bundled image into glance (in the xenapi code-path, that code lives within the xenapi glance plugin); from the sound of it, ewan wants to eventually pull that code back into nova proper, so eventually libvirt and xenapi will be able to share it16:31
vishyttx: not sure about that one.  I thought termie had a branch for it at one point, not sure what the status is.16:31
vishysirp-: I'm doing ultra simple snapshotting as a first pass, just using qcow2 format internal snapshotting16:32
vishysirp-: so no external backup or anything fancy like that16:32
sirp-vishy: ah gotcha16:33
*** kashyapc has quit IRC16:33
ttxtermie: if you have anything in the oven, let me know. Otherwise i'll just defer it to Cactus16:37
* soren dinners16:39
*** lvaughn_ has quit IRC16:52
*** lvaughn has joined #openstack16:53
*** Jordandev has joined #openstack16:54
*** zul has quit IRC16:56
*** DigitalFlux has quit IRC16:58
*** anotherjesse has joined #openstack17:00
*** dillon-w has joined #openstack17:04
*** nelson__ has quit IRC17:04
*** opengeard has quit IRC17:04
*** nelson__ has joined #openstack17:05
*** Jordandev has quit IRC17:06
*** dragondm has quit IRC17:12
johnpurvish: q on "directapi"?17:14
dillon-wcreiht around?17:15
*** cynb has joined #openstack17:20
*** Jordandev has joined #openstack17:20
Ryan_Laneinstance creation appears broken in trunk: TypeError: _format_instances() takes exactly 2 arguments (3 given)17:20
Ryan_Lanethough I'm using trunk ppa, which I guess could be out of date: 2011.1~bzr522-0ubuntu0ppa1~lucid117:21
notmynamedillon-w: go ahead and ask. he'll see it when he gets in or someone else can answer17:22
daboeday: Just pulled trunk, and noticed that internal_id has been removed from the instances table. Much of the python=cloudservers API is configured to use that value, not the database ID.17:24
daboeday: can you explain the reasoning for dropping it, and if it is expected that we only use the database id from now on?17:24
dillon-wI've got new swift-bench result.
dillon-wis it still too low?17:25
notmynameare you using dedicated hardware or a VM or ...?17:26
dillon-wdedicated h/w17:27
notmynamedillon-w == winston-d?17:27
*** rlucio has joined #openstack17:28
*** zul has joined #openstack17:29
dillon-wgood catch. that's me too.17:31
colinnichdillon-w notmyname To give another set of figures, my test system has a proxy on XenServer, and 3 storage nodes on physical hardware in different locations (one 400 miles away). Don't ask... Here are my swift-bench results -
dillon-wi'm now signing-in from home17:31
dillon-wcolinnich: that's not bad considering your setup.17:33
colinnichdillon-w: so your writes don't look fantastic compared to mine, but GETs look faster17:33
dillon-wcolinnich: you are right. but i am using 6 storage nodes in the same LAN17:34
colinnichdillon-w: so you should be much faster than me at everything I'd expect17:35
dillon-wcolinnich: i guess so.17:35
dillon-wi've spent two days debugging (well, asking help here) the slow perf issue17:36
colinnichdillon-w: can't really offer anything else, I haven't done any tuning at all.17:37
dillon-wcolinnich: i think there's something wrong in my swift config or setup, 'cos my storage device is super fast. :D and I confirm that with native IO mirco benchmarks17:38
edaydabo: we only need one ID to reference resources, and eventually this should be globally unique (whether a UUID or some other form)17:39
daboeday: that's great in theory, but it breaks a lot of what I'm working on now17:39
edaydabo: how was that being exposed by the API? from a client it shouldn't matter, no?17:40
daboeday: more of a question of timing - why this happened one day before the deadline17:40
edayjust whatever ID's are in the URL17:40
dillon-wwhat should i expect from 'swift-account-audit'?17:40
daboeday: no, the api code that handles requests assumes internal_id17:41
daboclient doesn't care17:41
edayand the thing you're working on just uses the REST API, doesn't it?17:41
dabono, it implements the REST API17:42
* eday is confused17:42
edayare you working on nova/api/openstack/* modules?17:42
*** lvaughn_ has joined #openstack17:43
daboeday: yes, primarily17:43
edayahh, ok. you mentioned python-cloudservers, so I thought you were focusing only on client17:43
daboeday: that's the most obvious place where it's breaking. A simple 'cloudservers list' is crashing now17:44
daboI'm also implementing the admin password reset blueprint17:44
*** lvaughn has quit IRC17:45
edayif you want to push your branch I can help convert, it should be a simple replace of internal_id with compute.api methods to use normal IDs17:45
daboeday: well, I'm already about halfway down that road. I just wanted to understand the reason why this change was made now.17:45
*** trin_cz has quit IRC17:45
edaydabo: trying to get API cleanup done before the merge req window closes17:46
daboeday: exactly - I'm trying to get my stuff merged, too17:47
edayit's a step in the distributed scheduler, as the previous internal_id geenration was kind of useless and only complicated having two IDs per resource17:47
edaydabo: sorry, conflicts are hard to avoid with so much parallel dev going on :)17:47
*** joearnold has joined #openstack17:47
daboeday: hey, I know it wasn't intentional. ;-)17:48
edaydabo: if it makes you feel any better, I had to merge trunk with that thing like 10 times because everyone kept submitting branches with conflicts (assuming old id's/compute API)17:49
dillon-wis there anything wrong with the 'swift-account-audit' result here ?17:49
daboeday: been there, done that. :)17:49
*** dillon-w has quit IRC17:55
*** Jordandev has quit IRC17:58
*** rogue780 has joined #openstack18:02
*** aliguori_ has quit IRC18:03
colinnichdillon-w: I'd certainly say so. As for the cause, I'm no expert. I'd guess checking all your services on the storage nodes are running, but other than that I'll leave it to the experts.18:04
*** jdurgin has joined #openstack18:08
*** anotherjesse has quit IRC18:10
*** Jordandev has joined #openstack18:11
Ryan_LaneAPI stuff in trunk seems pretty broken18:12
pvoRyan_Lane: dabo and eday are working on it now18:14
Ryan_Laneah. ok18:15
Ryan_Lanedabo, eday: incase you aren't already aware. display_name isn't set properly on instance creation18:15
Ryan_Lanefor some reason it is setting a default, even when display_name is passed in18:16
*** dragondm has joined #openstack18:16
daboRyan_Lane: what are you seeing vs. what are you expecting?18:16
*** maplebed has joined #openstack18:16
Ryan_LaneI'm passing in "test" as the display name, and it is setting "Server 36"18:16
Ryan_LaneI traced the code last night18:16
daboRyan_Lane: creating the instance with what? different tools do it differently18:17
Ryan_Lanein the compute api, it sends the info to the db layer, and then can't find display_name in the returned instance18:17
Ryan_LaneI'm having a good feeling this affects all apis18:17
Ryan_Lanesince after tracing the code, it makes it to the underlying code18:18
Ryan_LaneI'm using the aws-sdk18:18
daboRyan_Lane: right now I'm trying to clean up all the stuff that references instance 'name', since a lot of that is broken. I haven't looked at 'display name'18:19
*** dirakx has joined #openstack18:19
Ryan_Lanein compute/, in create, if I change:18:19
Ryan_Lane            if 'display_name' not in instance:18:19
Ryan_Lane                updates['display_name'] = "Server %s" % instance_id18:19
sandywalshdabo, are you looking at that or should I?18:20
Ryan_Laneupdates['display_name'] = display_name18:21
dabosandywalsh: if you want to start looking at display name, go for it. I'm knee-deep in cleaning up the name references18:21
Ryan_Laneif I set that, it works as expected18:21
sandywalshRyan_Lane, I'll look at that now18:21
Ryan_Laneso the display_name makes it into the function :)18:21
Ryan_Lanesandywalsh: thanks :)18:21
Ryan_Lanethat's breaking my development pretty hard. heh18:21
Ryan_Lanealso, the ec2 api can't create vms at all18:21
sandywalshI think that's a separate issue18:22
Ryan_Laneit is :)18:22
sandywalshI've been focused on the openstack api as of late18:22
Ryan_LaneI'm getting: _format_instances() takes exactly 2 arguments (3 given)18:22
rluciowhoa when did the db schema change?18:23
Ryan_Lanewhich is this line: i = self._format_instances(context, reservation_id)18:24
rluciofor nova i mean18:24
Ryan_Lane(I really need to switch to using the openstack api soon)18:24
vishyrun_instances is broken in trunk?18:24
Ryan_Lanefor ec2, yes18:25
*** jc_smith has joined #openstack18:25
vishyjohnpur: sorry I missed that comment, what was the q?18:25
vishyis there a bug/patch in?18:25
Ryan_Laneno. I can submit a bug though18:25
Ryan_Lanedidn't want to if it was a known issue :)18:26
Ryan_LaneI'll submit one for display_name as well18:26
*** westmaas has joined #openstack18:26
sandywalshRyan_Lane, I can put that my api-parity branch18:26
sandywalshRyan_Lane, seems the 'not in' is failing because Instance is a sqlalchemy object18:29
Ryan_Laneheh. that would make sense :)18:29
annegentlecould I get one more Nova Core review on this merge request? pretty please?18:29
Ryan_LaneI assumed functions were overloaded to make that happen18:30
sandywalshlikewise ... digging deeper18:30
Ryan_Lane(which is one of the reasons I don't really like overloaded operators)18:30
sandywalshheh, too ruby-like18:31
* sandywalsh ducks and covers18:31
Ryan_Laneit's confusing, and makes you have to search for something that may not be there18:31
*** dragondm has quit IRC18:32
vishyi have a fix18:32
vishyfor format18:33
sandywalshRyan_Lane, display_name was set correctly to start with. It gets overwritten18:33
vishyi'll make a bug and fix18:33
Ryan_Lanejust made a bug18:33
vishyfix proposal is here:
Ryan_LaneI'll test when the diff show up :)18:39
vishyi'm just looking at tests, looks like we have some broken ones18:40
*** aliguori has joined #openstack18:40
Ryan_Laneworks for me now18:40
uvirtbotNew bug: #698225 in nova "run_instances broken in trunk: problem with _format_instances passed too many arguments" [High,Fix committed]
dabovishy: doesn't fix _format_instances for me18:42
vishydabo: huh?18:43
vishyfixing tests as well18:44
daboI applied the diff, and still get the same error18:44
daboit looked like it should work18:44
vishydid you restart api?18:44
*** mdragon has joined #openstack18:44
dabomy bad18:44
Ryan_LaneprivateDnsName isn't returned on instance creation (EC2 API), but is returned when running describe_instances18:46
vishyRyan_Lane: true18:46
*** irahgel has left #openstack18:46
Ryan_Laneis that by design?18:46
vishyit is also how amazon does it18:47
*** mdragon has left #openstack18:47
vishyit doesn't get an ip until it has been scheduled to a host18:47
Ryan_Laneso I should create the instance, then do describe_instance using the instanceid.18:47
Ryan_Lanethat's problematic18:47
Ryan_LaneI'm assigning a DNS name on creation18:48
Ryan_LaneI guess I could loop describe_instance until I get an IP18:48
vishyRyan_Lane: you're doing dns externally?18:48
*** dragondm has joined #openstack18:49
vishyRyan_Lane: patch nova to do internal dns!18:49
Ryan_LaneI'm doing it via LDAP18:49
Ryan_LaneI can do that if you'd like18:49
Ryan_Laneit would make my life easier, for sure18:49
vishyRyan_Lane: nova should be doing dns of some sort18:49
Ryan_Lanecool. I'll add it there :)18:50
vishyRyan_Lane: ips are much more flexible if we assign them later18:51
*** kevnfx has quit IRC18:51
vishythere was a blueprint and such about it.18:51
Ryan_Laneyeah. I'd imagine so18:52
Ryan_Lanelet me check to see how portable the LDAP stuff is18:52
Ryan_Lanefor DNS18:52
Ryan_Laneit may not be worth adding if it only works with PowerDNS (which is what I'm testing)18:52
Ryan_LaneI guess I could make it abstract and add subclasses for specific server types, and have it a choice via a flag...18:54
jk0to anyone interested, I wrote up a bzr pre-commit hook that runs pep8 and unit tests:
jk0unit test checks are disbale by default since they take forever18:56
edayjk0: how long do they take for you?18:57
jk0~2 minutes18:58
edayyou can get it down under 1 min with libeatmydata :)18:58
* jk0 googles18:58
eday(it's a LD_PRELOAD to disable fsync/fdatasync)18:58
openstackhudsonProject nova build #363: SUCCESS in 1 min 16 sec:
edayit's due to sqlite being all persistent and stuff18:59
eday(the slowness)18:59
*** Jordandev has quit IRC18:59
sandywalshRyan_Lane, can't do 'foo' in object with sqlalchemy objects (only dicts). Looks like sqlalchemy objects create proper attributes and don't do any magic (which is nice).19:00
sandywalshRyan_Lane, putting patch in api branch19:00
Ryan_Lanesandywalsh: great :)19:01
sandywalshnp, it was on my list of things to look at anyway.19:01
rogue780ok. I've got a dumb question.19:06
*** adiantum has quit IRC19:06
rogue780I've installed nova and have an instance running. I can ssh into it from the host machine at
rogue780So my question what?19:07
rogue780how do I use that instance to, say, run a web server? and how do I access that instance from a client machine?19:08
*** Jordandev has joined #openstack19:11
creihtdoh... dillon-w is gone19:14
vishyjk0: LockContention: Could not acquire lock "/home/vishvananda/trunk/.bzr/checkout/dirstate": [Errno 11] Resource temporarily unavailable19:25
*** gdusbabek has left #openstack19:25
jk0that's a new one to me19:25
*** gdusbabek has joined #openstack19:26
vishyjk0: also, i think a warning would be better than an error, I don't want to have to make tests/pep pass to just commit19:26
vishyjk0: and there doesn't seem to be a way to do a pre-push hook19:26
jk0yeah, I don't think one exists19:26
jk0otherwise a pre-push would be preferable19:27
jk0I can make it a warning though19:28
vishyrogue780: assuming it is an ubuntu image, you can apt-get install apache219:28
vishyrogue780: you can also associate floating ips to the instance assuming you have extra ips to give19:28
vishyjk0: i thought maybe the LockContention was due to it running the test to check authors while in the middle of pre-commit which locks the dir, but you don't get that error?19:30
jk0no, I haven't seen that19:30
jk0a few other people have tested as well, too19:30
jk0-, too19:30
*** ccustine has quit IRC19:33
rogue780vishy, tell me more about these floating ips? I'm interested in making it so someone from the outside world can access the webserver on the instance...I have next to no knowledge of how to do this.19:33
*** zykes- has quit IRC19:36
*** Ryan_Lane is now known as Ryan_Lane|food19:37
*** hggdh has quit IRC19:38
*** hggdh has joined #openstack19:44
*** troytoma_ has joined #openstack19:50
creihtcolinnich: even given your setup, your bench results seem very sllow19:53
*** zul has quit IRC19:54
colinnichcreiht: :-(19:54
colinnichcreiht: I got some new machines today, so I'll have a proper set up in the next few weeks.......19:55
*** zykes- has joined #openstack19:55
creihtcolinnich: did you tune the workers any?19:55
*** dubsquared has joined #openstack19:55
colinnichcreiht: not at all, no19:56
creihtall the servers default to 1 worker, unless you specify it19:56
creihtshows the settings for each service19:57
colinnichcreiht: I'll give it a go on my current system then, will be good practice19:58
*** nelson__ has quit IRC19:58
*** nelson__ has joined #openstack19:59
*** joearnold has quit IRC20:00
sandywalshtr3buchet, thanks for the review20:04
sandywalshand YOU TOO could be as cool as tr3buchet by having a look at
daboas cool as tr3buchet? That's hardly motivation!20:05
sandywalshdunno dabo ... tr3buchet's got street cred20:05
*** littleidea has quit IRC20:05
daboyeah - sesame stree20:06
dabostreet, even20:06
sandywalshI think that illustrates my point ;)20:06
colinnichcreiht: a little faster, not too much. That's turning proxy-server up from 1 to 820:10
colinnichcreiht: object,container and account servers are at 2 - should that be higher?20:10
vishyrogue780: they work like amazon's elasticips.20:21
vishyrogue780: you will have to specify 1 or more floating ips with nova-manage floating create20:22
vishyrogue780: then using euca2ools euca-allocate-address, euca-associate-address,20:22
*** aliguori has quit IRC20:24
Ryan_Lane|foodvishy: I was wondering about floating ips as well20:25
*** Ryan_Lane|food is now known as Ryan_Lane20:25
Ryan_Lanehow does it assign the IP to the vm?20:25
Ryan_Laneis it via NAT?20:26
Ryan_Laneand is it available in flat networking mode?20:26
colinnichcreiht: turning the storage server services up to 8 gave another marginal improvement. I don't think it's worth reading too much into it though - the furthest server is around 60ms away20:27
edaysandywalsh: why are you hashing the image_id in your api-parity branch? What about doing a two-way conversion to something that is acceptable?20:30
sandywalsheday, the hash was first thing that came to mind if you have a two way function, that would be excellent.20:31
edaysandywalsh: what do you need for output format?20:32
sandywalshstring <-> number20:32
edayhm, and object store image ID's are pretty much base64, no? would really big ints be ok? :)20:33
sandywalshit would have to be a perfect hash20:33
edayoh, image id's have '-' don't they... hmm20:34
sandywalsheday, yeah, but glance is numeric20:34
edayunless you could always rip off 'ami' prefix20:34
sandywalshwe could do that ... is ami fixed?20:34
edayif you ripped off 'ami-' and took the other bits and converted to int via base16/36/64/whatever fits the best, might be ok20:35
edayjust the 'grab all and compute hash until one matches' loop seems nasty :)20:35
sandywalshwell, for that matter I could just mod the existing hash to 1/2 the size20:35
sandywalshthe loop sucks20:36
sandywalshbut with glance I can just request by id20:36
*** fabiand_ has joined #openstack20:36
sandywalsheday, it's more of an encryption function than a hash that's needed20:37
vishyRyan_Lane: it is via nat and it should work in flat mode as well20:43
Ryan_LaneI see it is assigned per compute node...20:43
Ryan_Lanefuture version of nova are going to allow instance migration, correct?20:44
Ryan_Lanewill floating IPs move between compute nodes with the instance?20:44
vishyRyan_Lane: yeah there is a live migration branch in for review, but I think it requires NAS20:44
Ryan_Lanenetwork attached storage?20:44
Ryan_Lane(I'm going to have that :) )20:45
Ryan_Laneso i can have a private address range that auto-assigns IPs to instances, and have a public address range that is added via floating IPs?20:46
vishyRyan_Lane: yes, that is how we deploy20:46
vishyRyan_Lane: no public access by default20:46
Ryan_Laneis there documentation on how to set this up anywhere?20:47
vishyprobably not :)20:47
annegentleheh, yes I'm reading with interest myself :)20:49
Ryan_Lanethis is definitely something I need to set up. we're really low on IPv4 ip addresses20:50
Ryan_Laneand we can't get any more20:50
*** allsystemsarego has quit IRC21:03
sandywalshJust did a fresh merge with trunk and now 'boot' fails in os api. eday dabo ?21:08
nelson__ <--- Perl and Python programmers compared.21:09
*** joearnold has joined #openstack21:09
sandywalshinteresting also to note that the instance id sent to "cloudservers list" was 4, vs 334743 (or whatever) eday dabo21:10
*** aliguori has joined #openstack21:10
sandywalsh eday dabo21:11
nelson__I'm getting this error and I'm not sure where to look to find a solution: ClientException: Object PUT failed: 404 Not Found21:14
nelson__Should I be looking at the proxy or at one of the storage servers?21:14
dabosandywalsh: I'm removing all the references to 'internal_id' out of the code21:14
daboI won't be ready to push for a while, though21:15
notmynamenelson__: 404 means that the account, container, or object could be missing21:15
*** Jordandev has quit IRC21:15
notmynamefor a PUT, I'd guess it's the container21:15
nelson__notmyname: the instructions don't seem to tell me to create the container first.21:15
creihtnelson__: start with the proxy server logs for the request, which will have a transaction id (start with tx and then a uuid), then grep all the logs for that transaction id21:15
*** Jordandev has joined #openstack21:15
creihtst is supposed to create the container if it doesn't exist (if that is what you are using)21:16
notmynamewhich instructions are you using?21:16
creihtI'm guessing the multi-server instructions21:16
nelson__yes, multi-server, and yes, I'm using st, and no, I'd look in /service/swift-proxy/logs/main but of course I'm not using daemontools, so my log output could be just about anywhere.21:17
creihtnelson__: can you do something like st ARGS stat AUTH_0cbd1929b0d94dc196d90620095657ea21:18
creihtwhich will make sure it can find the account21:18
creihtnelson__: by default it logs to syslog or messages (depending on distro)21:18
nelson__creiht: that container isn't found.21:19
creihtoh sorry21:19
creihtjust st ARGS21:19
creihtwill check the account21:19
creihtst ARGS stat21:20
nelson__0 containers 0 objects 0 bytes21:20
nelson__so it's definitely not even creating the container21:20
creihtbaby steps21:21
creihtso the account exists21:21
creihtnow do st ARGS stat myfiles21:21
nelson__Container 'myfiles' not found21:22
sandywalshdabo, can you ping me when you push? broken branch.21:24
dabosandywalsh: sure21:24
*** ccustine has joined #openstack21:25
*** fabiand_ has quit IRC21:26
creihtnelson__: ok so we verified that the container didn't get created21:28
nelson__yep, and I can't find log output from swift-proxy-server21:28
creihtnelson__: which os?21:29
nelson__Ubuntu 10.04.121:29
creihtthen they should be in /var/log/syslog, unless you made any logging changes21:30
creihtso if you grep syslog for "proxy" you don't find anything?21:31
nelson__I restarted it; now I do.21:31
creihtrestarted syslog? or the proxy?21:31
nelson__the proxy.21:32
uvirtbotNew bug: #698316 in glance "Glance reads the whole image into memory when handling a POST /images request" [Undecided,In progress]
nelson__it's consulting the auth, then account, then proxy-server21:32
creihtso the easiest thing might be to tail syslog and run the st command again21:32
nelson__hrm. the account server is returning the 404.21:33
creihtare you sure you are sending the correct account string?21:33
nelson__and yet st ... stat is returning information.21:33
nelson__am following the instructions from installmulti, which says this: st -A -U system:root -K testpass stat21:34
nelson__which I previously successfully (?) created.21:35
dabocan anyone more familiar with eventlet looping help me avoid this? The task is timing out on the remote, so I want to exit at that point.21:35
uvirtbotdabo: Error: "^" is not a valid command.21:35
dabo  ^^
creihtdabo: an eventlet Event can only be sent a value once21:36
dabocreiht: Yeah, I'm only calling it once.21:36
daboI'm trying to figure out why raising an exception calls eventlet a second time21:37
creihtdabo: can you paste the surrounding code where the error is happening?21:37
creihtnelson__: can you paste the proxy log with the 404 and surrounding swift lines?21:38
dabok - gimme sec21:38
sandywalshdabo, it's the LoopingCode stuff that reuses the Event incorrectly on exception.21:39
sandywalshI haven't had a chance to dig any deeper21:39
jk0can I get a couple reviews for please? it's just a ~5 line change21:39
nelson__Jan  6 21:31:15 ubuntu auth-server - - [06/Jan/2011:21:31:15 +0000] "GET /v1.0 HTTP/1.0" 204 - "-" "-" - - - - - - - - - "-" "" "-" 0.002321:39
nelson__Jan  6 21:31:15 ubuntu account-server - - [06/Jan/2011:21:31:15 +0000] "HEAD /sdc1/211931/AUTH_0cbd1929b0d94dc196d90620095657ea/myfiles" 404 - "txca9791fb-683c-4236-a29a-05378ba11a07" "-" "-" 0.0012 ""21:39
nelson__Jan  6 21:31:15 ubuntu proxy-server - 06/Jan/2011/21/31/15 HEAD /v1/AUTH_0cbd1929b0d94dc196d90620095657ea/myfiles HTTP/1.0 404 - - system%3Aroot%2CAUTH_tk8b2ad4f69572449ab1c48e257ad68102 - - - txca9791fb-683c-4236-a29a-05378ba11a07 - 0.063121:40
dabosandywalsh: yeah - the question is why?21:40
creihtnelson__: please21:40
nelson__it's only three lines.21:40
nelson__but as you wish21:40
*** daleolds has joined #openstack21:40
*** opengeard has joined #openstack21:42
creihtnelson__: ok the account-server is returning the 404 on the existence of the myfiles container21:42
*** ctennis has quit IRC21:43
nelson__but we already know that it doesn't exist, right?21:43
*** miclorb has joined #openstack21:44
nelson__why is the account server being asked about containers? Shouldn't the container server be asked??21:44
creihtnelson__: I'm going to try to reproduce here21:45
creihtwell I was a little wrong there21:45
creihtbut I need to double check21:45
nelson__I think I'm going to put these log messages into the install docs, as in "on success you'll see this".21:45
nelson__that will also get installers to find the log files so they can self-help better.21:45
uvirtbotNew bug: #698326 in nova "Trunk breaks diagnostics API query" [Undecided,In progress]
creihtnelson__: what does your st command look like that you are using to upload?21:49
nelson__straight out of the documentation :) so if it's wrong, I fixy21:49
creihtnelson__: for grins sake, will you please just show me the command you are running?21:50
nelson__st -A -U system:root -K testpass upload myfiles bigfile1.tgz21:50
creihtfor some reason on the head in your code it is attaching /myfiles to the account21:51
creihtit doesn't do that when I run it here21:51
nelson__hrm.  bug in 1.1?21:51
creihtso I wanted to see if maybe a / got in there accidently21:51
creihtnelson__: don't know yet21:52
creihtyou are running 1.1?21:52
*** bvierra has quit IRC21:52
annegentlenelson__: go go go :)21:52
nelson__I'm following the instructions, which I believe lead me to install 1.121:52
creihtnelson__: I'm just trying to verify to make sure that I test the right thing21:53
creihtwhen debugging issues, I don't like to assume anything21:53
nelson__yes, it seems to be installing from ppa:swift-core/ppa21:53
dabosandywalsh: Pushed my changes: lp:~ed-leafe/nova/xs-password-reset21:54
*** Xenith has quit IRC21:54
*** ctennis has joined #openstack21:55
*** ctennis has joined #openstack21:55
sandywalshdabo, but I need changes from trunk21:56
sandywalshtrunk is busted21:56
daboI merged in trunk a short while ago21:56
daboI do that at least once a day21:56
sandywalshbut then my branch will have all your password reset stuff21:57
daboyou could diff the two and ignore any of the password stuff21:57
sandywalshwe can get your patch merged back into trunk faster if it was a separate branch21:57
sandywalshlet's get trunk fixed21:58
creihtnelson__: I haven't been able to reproduce it yet, even from the 1.1.0 tarball :/21:58
creihtI have to go to a meeting, but will look again in a moment21:58
dabothen you'll have to wait a while. I've been staring at this monitor for almost 12 hours now. I need a break21:58
nelson__creiht: thanks.21:58
sandywalshdabo, gotcha21:59
nelson__creiht: note that this is upload, not post.22:01
creihtright, I'm testing by doing the exact same thing you are22:01
*** Xenith has joined #openstack22:01
*** kevnfx has joined #openstack22:02
*** miclorb has quit IRC22:05
*** daleolds has quit IRC22:06
uvirtbotNew bug: #698336 in nova "Insufficient resources (memory) causes instance to pending state forever" [Undecided,New]
*** trin_cz has joined #openstack22:06
*** miclorb has joined #openstack22:13
*** littleidea has joined #openstack22:17
*** spectorclan has quit IRC22:19
*** rogue780 has quit IRC22:23
*** damon__ has joined #openstack22:25
*** jdarcy has quit IRC22:30
dabook, sandywalsh, eday and anyone else: this merge proposal should fix the issues in trunk due to removal of internal_id from instances:
vishydabo: there are some odd additions there22:37
dabovishy: such as? I didn't start with a clean branch; I'm trying to merge only some of what is in my branch22:38
vishydabo: you added back in get_by_internal_id?22:38
dabodid I screw it up?22:38
daboI thought I removed that22:38
vishyjust looking at the diff22:39
vishydabo: also, why did you change get back to get_instance?22:39
dabobecause it is called elsewhere, and get_instance was a clearer name22:40
termievishy, dabo: 'cause it's a great idea :)22:40
vishydabo: we should rename compute to instance :P22:40
dabovishy: yeah, looks like I missed the merge in nova/db/api.py22:40
dabolet me fix that22:40
vishydabo, termie: i don't really care i just don't want to be switching back and forth every other patch22:41
dabovishy: it was get_instance before22:41
termievishy: the underlying names are still get_instance22:41
termieon the manager22:41
termiei wanted to write a patch to change all the names back to their originals22:42
*** hadrian has quit IRC22:43
dabovishy: ok, pushed the fix22:43
dabogotta run out for a little while; back soon22:47
sirp-ran into an interesting little issue with eventlet-and-exceptions, for your edification:
termiesirp-: interesting, hadn't run into that before22:49
termiesirp-: by "uses_eventlet" what do you mean however?22:49
termiesirp-: all of our code 'uses' eventlet22:49
termiesirp-: do you mean uses eventlet specific commands like... greenthread.spawn(...)22:50
sirp-in this case, i the code was using httplib (which has been monkey-patched by eventlet via socket) to write out a message to another server22:50
sirp-ill make another paste, one sec22:50
edaytermie, dabo, vishy: as you saw, the reason for _instance removal was not just get, but all methods in compute.api. it was inconcsistent before22:51
creihtnelson__: and also just to verify, you are using the st that was installed via the package right?22:51
edaythere were more without, so I removed them (and it was needed before, since there wasn't the compute context)22:51
*** ppetraki has quit IRC22:51
edaywe can put em back on, but we should do it to all methods, not just a couple22:51
termieeday: i would prefer that, yeah22:52
vishyeday, termie, dabo: agreed, let's leave them as get and change everything in one patch22:52
termiesirp-: right, interesting -shrug- i don't know why it would do that22:53
termiesirp-: it tries to track exceptions when they happen22:53
termiebut i don't see why it would clear thins before hand22:53
sirp-termie: dragon was mentioning that it probably keeps exceptions from leaking between co-routines22:53
termiesirp-: aye, i just don't see why it would leak, you already know which coroutine you are in22:53
termiesirp-: but i suppose it could somehow22:54
*** Jordandev has quit IRC22:54
edayalthough if that's the move, I still think volume_api.create_volume() is redundant ;)22:54
sirp-termie: right, it probably just made it easier for them to code, and they may not have run into this quite yet22:54
*** Jordandev has joined #openstack22:54
nelson__creiht: yes, installed via the package22:55
termieeday: _api is redundant too, but it helps people keep track of what is going on a lot easier22:55
creihtnelson__: I'm trying everything that I can think of, but I can't reproduce the issue :/22:55
termieand in general i think it is worthwhile to have unique names across methods even if nothing flattens them22:55
nelson__creiht: even if you start with empty everything?22:56
nelson__I don't *think* I've misconfigured everything. Have cut-and-pasted from the instructions22:56
creihtseveral people have followed those instructions and haven't seen the same problem that you have found22:56
edaytermie: hehe. I thought about removing the API class in my last patch and just doing nova.volume.create() too :)22:56
creihtI'm not working on a multi-server install, but my saio22:57
jk0would anyone be so kind as to look over please?22:57
jk0it's tiny22:57
termiejk0: commented23:00
*** westmaas has quit IRC23:01
creihtnelson__: will you paste the reslt of running23:02
creihtsqlite3 /etc/swift/auth.db "select * from account"23:02
nelson__looks correct to my untutored eye.23:03
creihtverifing that an admin account was created23:04
creihtif it wasn't admin, then you wouldn't have perms to create the folder23:04
jk0termie: responded23:05
termiejk0: why not just leave it as datetime? it should be serialized later on, not by the db api23:06
termiejk0: everything else returns the datetime itself23:06
termiebut that change doesn't change whatever is consuming the output23:06
jk0it was raising exceptions about not being able to serialize the datetime object23:06
termiewell that is something to fix in the serialization library, see my easy_api branch for an example of how23:07
termiethe db api should return the proper info, the serialization layer should deal with it, datetimes are serializable you just need to touch them first23:08
*** daleolds has joined #openstack23:08
daboeday, vishy, termie - I've pushed a commit that changes 'get_instance' back to 'get'. I'd love to argue the merits, but we're facing a deadline, and this is holding lots of stuff up23:08
termiejk0: the _to_primitive is the function to look at23:08
jk0alright, I'll look at that23:08
edaydabo: yup, we can change em all or nothing later23:08
vishyjk0: if you only need second granularity on serialize unserialize there is also utils.isotime() and utils.parse_isotime()23:09
vishydabo: waiting for bzr to register the change23:10
dabovishy: yeah, the push seemed to take unusually long this time23:11
edayusually doesn't. push to the right branch name?23:11
vishydabo: it appears something strange happened to that branch23:14
* termie jumps in and motions to the cameras23:15
edaydabo: btw, I did like the _cast_compute_message stuff, something I was planning on doing today actually :)23:15
*** gondoi has quit IRC23:15
jk0vishy: do it would be acceptable to use utils.isotime() in that dict?23:15
termiejk0: i do not23:16
vishyjk0: I was referring to using it for serialization23:16
jk0oh ok23:16
jk0figured, just making sure23:16
dubsquaredanyone familiar with this issue when launching an UEC image:
vishydubsquared: your curl looks like it failed ami-7mxsmeqt/image23:23
vishyit either doesn't exist or objectstore is down23:23
vishy6      Couldn't resolve host. The given remote host was not resolved.23:24
vishydubsquared: no route to ?23:24
dubsquaredthats localhost :/23:25
dubsquaredtcp        0      0  *               LISTEN      7373/python23:25
vishyyou could try the curl by hand with --silent turned off and --verbose on to see why it is failing23:25
vishyperhaps you have routing conflict between instances and host?23:26
daboeday: I hate copy/paste code23:26
dabovishy: uh oh - what now?23:26
dubsquaredcurl: (22) The requested URL returned error: 500 —> internal server error…the best code ever!23:27
*** hggdh has quit IRC23:30
*** hggdh has joined #openstack23:31
*** troytoma_ has quit IRC23:31
JordanRinkevish, I have setup now, you can dump images there to be pushed out to the CDN and distributed for the community23:35
dubsquaredhmm, updated the packages on a working controller…then i can't launch images anymore23:35
JordanRinkeI am working on a format to enable automatic updating etc, but for now it is at least a blind repo23:35
dubsquaredsomething went awry23:35
termiedubsquared: sleepson- may have some insight, i think he was doing something like this yesterday23:35
vishydubsquared: check for error messages in objectstore log.23:36
vishydubsquared: you didn't pass in creds. in headers in your curl23:43
creihtnelson__: can you try restarting memcache on your proxy, and then try again?23:44
nelson__same error.23:45
creihtyeah that was reaching a bit23:45
creihtjust wanted to check if memcache had gotten some weird value in it23:45
dubsquaredvishy: they are all there, everything checked out…just pasted the exception23:47
creihtnelson__: is the proxy/auth on a separate server from the rest of the storage nodes?23:48
vishydubsquared: hmm, perhaps the user you are using doesn't have permissions for that image23:48
nelson__I have a storage server running on the proxy/auth ... but I can turn it off if you think that's causing a problem.23:48
creihtok just wanted to verify23:49
creihtthat should be theoretically fine23:49
dubsquaredusing the admin user atm, haven't messed with the other roles yet23:49
jk0termie: how about now?23:52
*** ctennis has quit IRC23:54
*** dirakx has quit IRC23:56
vishydubsquared: it looks like the underlying exception is getting eaten in some odd way23:58
vishydubsquared: in your paste it is printing out as None23:58
vishyline 137 in nova/objectstore/handler.py23:59
dubsquaredyeah, it is odd that the instances are being named as such:   DEBUG:root:instance 9: starting...23:59
dubsquaredthat's from compute.log..23:59

Generated by 2.14.0 by Marius Gedminas - find it at!