| *** adalbas has joined #openstack-dev | 00:30 | |
| *** adalbas has quit IRC | 00:36 | |
| *** jeblair has joined #openstack-dev | 00:53 | |
| *** littleidea has joined #openstack-dev | 01:12 | |
| *** pixelbeat has quit IRC | 01:36 | |
| hugokuo2 | morning | 01:54 |
|---|---|---|
| *** ywu has quit IRC | 02:51 | |
| *** andresambrois has joined #openstack-dev | 02:56 | |
| *** andresambrois is now known as aa | 02:56 | |
| *** rkukura has quit IRC | 03:18 | |
| nvez | If I want to upload a draft, who should I add in Gerrit for them to check? | 03:34 |
| *** shang has joined #openstack-dev | 03:35 | |
| *** edygarcia has joined #openstack-dev | 03:39 | |
| *** bhuvan_ has joined #openstack-dev | 03:47 | |
| *** edygarcia has quit IRC | 04:10 | |
| *** edygarcia has joined #openstack-dev | 04:23 | |
| *** dtroyer is now known as dtroyer_zzz | 04:26 | |
| *** edygarcia has quit IRC | 04:29 | |
| *** ncode has joined #openstack-dev | 04:30 | |
| *** ncode has joined #openstack-dev | 04:30 | |
| *** novas0x2a|laptop has quit IRC | 04:31 | |
| *** littleidea has quit IRC | 04:33 | |
| *** sdake has joined #openstack-dev | 04:39 | |
| *** garyk has quit IRC | 04:41 | |
| *** dachary has joined #openstack-dev | 04:42 | |
| *** vincentricci has joined #openstack-dev | 04:50 | |
| *** eglynn_ has joined #openstack-dev | 04:53 | |
| *** eglynn has quit IRC | 04:53 | |
| *** dachary has quit IRC | 04:59 | |
| *** eglynn__ has joined #openstack-dev | 04:59 | |
| *** dachary has joined #openstack-dev | 04:59 | |
| *** ncode has quit IRC | 04:59 | |
| *** eglynn_ has quit IRC | 05:01 | |
| *** hugokuo has joined #openstack-dev | 05:05 | |
| *** nvez is now known as mnaser | 05:07 | |
| *** hugokuo2 has quit IRC | 05:08 | |
| *** anderstj has joined #openstack-dev | 05:16 | |
| *** shang has quit IRC | 05:32 | |
| *** SQLStud has joined #openstack-dev | 05:34 | |
| *** SQLStud has left #openstack-dev | 05:39 | |
| *** trapni has joined #openstack-dev | 05:46 | |
| * trapni waves | 05:46 | |
| *** davidha has joined #openstack-dev | 05:47 | |
| *** mrunge has joined #openstack-dev | 05:58 | |
| *** garyk has joined #openstack-dev | 05:59 | |
| *** pmezard has joined #openstack-dev | 06:01 | |
| *** hattwick has quit IRC | 06:01 | |
| *** mindpixel has joined #openstack-dev | 06:06 | |
| *** anderstj has quit IRC | 06:07 | |
| *** vincentricci has quit IRC | 06:09 | |
| *** dachary has quit IRC | 06:12 | |
| *** hugokuo1 has joined #openstack-dev | 06:15 | |
| *** hugokuo has quit IRC | 06:19 | |
| *** longhudou has joined #openstack-dev | 06:30 | |
| *** reidrac has joined #openstack-dev | 07:07 | |
| *** jakedahn is now known as jakedahn_zz | 07:21 | |
| *** dachary has joined #openstack-dev | 07:21 | |
| *** berendt has joined #openstack-dev | 07:29 | |
| *** Stackops-Jorge has joined #openstack-dev | 07:31 | |
| *** davidha has quit IRC | 07:33 | |
| *** sdake has quit IRC | 07:39 | |
| *** apevec has joined #openstack-dev | 07:55 | |
| *** hattwick has joined #openstack-dev | 07:58 | |
| *** berendt1 has joined #openstack-dev | 08:02 | |
| *** bhuvan_ has quit IRC | 08:02 | |
| *** berendt has quit IRC | 08:02 | |
| *** davidha has joined #openstack-dev | 08:04 | |
| *** mancdaz has joined #openstack-dev | 08:04 | |
| *** pixelbeat has joined #openstack-dev | 08:08 | |
| *** Chuanyong has joined #openstack-dev | 08:16 | |
| *** darraghb has joined #openstack-dev | 08:18 | |
| *** Chuanyong has quit IRC | 08:18 | |
| *** Chuanyong has joined #openstack-dev | 08:18 | |
| *** fc__ has joined #openstack-dev | 08:25 | |
| *** azbarcea has joined #openstack-dev | 08:36 | |
| *** journeeman has joined #openstack-dev | 08:44 | |
| journeeman | Is there any other by which I could query which extensions a Quantum plugin supports apart from the plugin mentioning it in the list, supported_extension_aliases? | 08:46 |
| *** Ryan_Lane has quit IRC | 08:53 | |
| *** davidha has quit IRC | 08:56 | |
| *** davidha has joined #openstack-dev | 08:56 | |
| *** davidha_who_took has joined #openstack-dev | 09:08 | |
| *** davidha has quit IRC | 09:08 | |
| *** danpb has joined #openstack-dev | 09:11 | |
| *** GheRivero has joined #openstack-dev | 09:21 | |
| *** shang has joined #openstack-dev | 09:32 | |
| *** mancdaz has quit IRC | 09:50 | |
| *** mancdaz has joined #openstack-dev | 09:50 | |
| *** hugokuo1 has quit IRC | 09:56 | |
| *** hugokuo has joined #openstack-dev | 10:00 | |
| *** mcclurmc_ has joined #openstack-dev | 10:11 | |
| *** mjfork has joined #openstack-dev | 10:16 | |
| *** longhudou has quit IRC | 10:18 | |
| *** koolhead17 has quit IRC | 10:35 | |
| *** berendt1 has quit IRC | 10:42 | |
| *** Mkenneth has joined #openstack-dev | 10:55 | |
| davidha_who_took | Hi | 11:10 |
| *** davidha_who_took is now known as davidha | 11:11 | |
| davidha | In swift and in openstack in general, can one run multiple instances of teh same component on teh same server node? | 11:13 |
| *** milner has joined #openstack-dev | 11:14 | |
| davidha | E.g. run multiple Container Servers all serving teh same node (unlike SAIO example where different container servers serve different 'virtual' nodes) | 11:14 |
| davidha | I.e. they would all have the same "devices" config parameter (yet different "bind_port" config parameter) | 11:17 |
| *** journeeman has quit IRC | 11:25 | |
| *** azbarcea has quit IRC | 11:30 | |
| *** andresambrois has joined #openstack-dev | 11:45 | |
| *** bsza has joined #openstack-dev | 11:47 | |
| *** aa has quit IRC | 11:49 | |
| *** mrunge has quit IRC | 11:51 | |
| *** heyho has joined #openstack-dev | 11:55 | |
| *** salgado has joined #openstack-dev | 12:03 | |
| *** salgado has quit IRC | 12:03 | |
| *** salgado has joined #openstack-dev | 12:03 | |
| *** apevec has quit IRC | 12:03 | |
| *** markvoelker has joined #openstack-dev | 12:09 | |
| *** alaski has joined #openstack-dev | 12:10 | |
| *** rbasak has joined #openstack-dev | 12:27 | |
| *** ogelbukh has quit IRC | 12:29 | |
| *** akscram has quit IRC | 12:30 | |
| *** Mkenneth has quit IRC | 12:30 | |
| *** ogelbukh has joined #openstack-dev | 12:30 | |
| *** akscram has joined #openstack-dev | 12:30 | |
| *** berendt has joined #openstack-dev | 12:34 | |
| *** markmc has joined #openstack-dev | 12:34 | |
| *** koolhead17 has joined #openstack-dev | 12:34 | |
| *** alaski has quit IRC | 12:35 | |
| *** alaski has joined #openstack-dev | 12:36 | |
| *** dolphm has joined #openstack-dev | 12:36 | |
| *** dprince has joined #openstack-dev | 12:37 | |
| *** adalbas has joined #openstack-dev | 12:45 | |
| russellb | garyk: hello! so i hear you're working on quantum rpc :) | 12:45 |
| garyk | russellb: ping | 12:45 |
| russellb | what's the current issue? | 12:46 |
| garyk | russellb: hi, yes, i have encountered some problems (which i am certain are in the code that i have writte). i have a rpc.cast from the agent to the plugin. | 12:46 |
| garyk | on the plugin i do not see the method "callback" being called. | 12:47 |
| *** sdake has joined #openstack-dev | 12:47 | |
| garyk | i captured the traffic and see that the TCP connection receives the ACK's for the cast. | 12:47 |
| russellb | ok, so let's take a look at 2 lines of code ... the code doing create_consumer() on one side, and the rpc.cast() on the other. can you paste those somewhere? | 12:47 |
| *** sdake has quit IRC | 12:48 | |
| russellb | paste.openstack.org | 12:48 |
| russellb | or your pastebin of choice | 12:48 |
| *** sdake has joined #openstack-dev | 12:48 | |
| garyk | russellb: tx. in 2 | 12:48 |
| russellb | k | 12:48 |
| garyk | russellb: http://fpaste.org/Chwk/ (please note i had to chnage the rpc.register - this was due to the fact that quantum does not make use of a large amount of common libraries from openstack) | 12:51 |
| garyk | russellb: the rpc backend is qpid | 12:54 |
| russellb | i would s/context, value/self, context/ | 12:54 |
| russellb | but that shouldn't cause an error at this point | 12:54 |
| russellb | i think the code might require args in the message, even if you don't have any args so ... {'method': 'hello_world', 'args': {}} | 12:55 |
| russellb | if that was the problem, i'd expect an exception on the receiving side though | 12:56 |
| russellb | you have logging on the receiving side? | 12:56 |
| garyk | russellb: did not solve. no, i do not have logging enabled. . i'll give that a bash. | 12:57 |
| *** derekh_ has joined #openstack-dev | 12:58 | |
| *** derekh_ is now known as derekh | 12:59 | |
| russellb | ok, let me know if you get anything in the log | 12:59 |
| *** dolphm has quit IRC | 13:00 | |
| *** blamar has joined #openstack-dev | 13:01 | |
| garyk | russellb: ++1 http://fpaste.org/AF1X/ | 13:03 |
| russellb | ah ha | 13:03 |
| russellb | from openstack.common import local | 13:03 |
| *** Mkenneth has joined #openstack-dev | 13:03 | |
| garyk | russellb: great thanks! | 13:04 |
| russellb | np, let me know what happens after you get the local piece in there | 13:05 |
| russellb | just need to copy that file in from openstack-common | 13:05 |
| *** yazirian has quit IRC | 13:05 | |
| garyk | thanks. at the moment i have copied all of the necessary files to a openstack/common directory that i have. at the moment this is part of the quantum code. it would certainly be better to import it from the common openstack project. i understand that the specifics of this library are stil in debate. | 13:06 |
| *** yazirian has joined #openstack-dev | 13:06 | |
| russellb | garyk: right now each project is just copying in the pieces that they need | 13:07 |
| russellb | there is a script in openstack-common to help you copy in stuff to keep it up to date ... and a config file in each project that says which bits to copy in | 13:07 |
| *** littleidea has joined #openstack-dev | 13:07 | |
| garyk | russellb: ok. this is what i have done. the problem is that we have to tweak the code. i have yet to understand this | 13:07 |
| russellb | you shouldn't have to tweak anything that comes from openstack-common. if you do, we should talk about it | 13:09 |
| russellb | you'll have to tweak rpc, because it's not ready for common yet | 13:09 |
| russellb | o | 13:09 |
| russellb | i'll probably work on that some more this week | 13:09 |
| *** dolphm has joined #openstack-dev | 13:10 | |
| garyk | russellb: yup, this is what i have done. i made two chnages - the first was just updating the imports so that they are not using nova and the second was just to chnage the api for the register function (quantum did not use the cfg format) | 13:12 |
| *** cdub has joined #openstack-dev | 13:12 | |
| russellb | ok. if you create an openstack-common.conf and use the update.py script from openstack-common to do the copying for you, it will set up the imports for you automatically | 13:13 |
| *** kbringard has joined #openstack-dev | 13:13 | |
| garyk | russellb: great! would have saved me a few hours :) | 13:14 |
| russellb | heh, sorry | 13:14 |
| russellb | take a look at the .conf file in nova as an example | 13:14 |
| garyk | russellb: i should have asked. | 13:14 |
| garyk | russellb: can you please take a peak at - http://fpaste.org/ONxD/ - exception solved - still no callback | 13:14 |
| cdub | garyk: ping | 13:15 |
| *** roge has joined #openstack-dev | 13:16 | |
| garyk | cdub: pong | 13:16 |
| trapni | my nova.projects table got corrupted just within the last 1-2 hours. this happened weeks ago, too. what can cause this? | 13:16 |
| *** asalkeld is now known as asalkeld_zzz | 13:16 | |
| *** mancdaz has quit IRC | 13:16 | |
| cdub | garyk: i vaguely recall you having an issue w/ dnsmasq not responding some time back | 13:17 |
| trapni | since there're only two sql tables being affected, I don't think of any disk I/O error. | 13:17 |
| *** mancdaz has joined #openstack-dev | 13:17 | |
| *** Mkenneth has quit IRC | 13:17 | |
| *** sandywalsh has joined #openstack-dev | 13:18 | |
| garyk | cdub: correct. i'll check my sent mail to see if i can find it. | 13:18 |
| cdub | garyk: thanks | 13:18 |
| *** hattwick has quit IRC | 13:21 | |
| *** dtroyer_zzz is now known as dtroyer | 13:23 | |
| garyk | cdub: on fedora 17 I had problems with a dormant: dnsmasq process and Additional instance launching failed: (nova.rpc.common): TRACE: Stderr: '\ndnsmasq: failed to create listening socket for 10.0.0.1: Address already in use\n' | 13:24 |
| garyk | . in addition to this i saw: https://bugs.launchpad.net/nova/+bug/953712 and https://bugs.launchpad.net/nova/+bug/953708. | 13:24 |
| uvirtbot | Launchpad bug 953712 in nova "force_dhcp_release is currently unsupported on Fedora" [Undecided,Invalid] | 13:24 |
| russellb | garyk: take a look in rpc/amqp.py, ProxyCallback.__call__ | 13:24 |
| russellb | garyk: we are seeing the "received ..." log message, and the one that comes from unpack_context() | 13:25 |
| cdub | garyk: hmm, ok...not what's happening here. i'll dig in. thanks | 13:25 |
| russellb | so maybe there is a problem when it tries to hand it off to the thread pool | 13:25 |
| russellb | maybe the thread pool is busted? | 13:25 |
| russellb | is the code doing eventlet.monkey_patch() ? | 13:26 |
| russellb | it needs to | 13:26 |
| garyk | russellb: i'll check | 13:26 |
| russellb | when it first starts up | 13:26 |
| garyk | cdub: i think that killing the process was a work around. | 13:26 |
| russellb | all of the nova binaries do "import eventlet ; eventlet.monkey_patch()" as the very first thing, before importing anything else | 13:27 |
| cdub | garyk: yeah, that's not the issue here | 13:27 |
| koolhead17 | hi all | 13:27 |
| garyk | russellb: no the code did not call eventlet.monkey_patch() [is this my monkey business?] | 13:29 |
| russellb | :) | 13:30 |
| russellb | try adding that ... | 13:30 |
| garyk | ok, will do | 13:30 |
| *** Mkenneth has joined #openstack-dev | 13:31 | |
| *** mestery has quit IRC | 13:33 | |
| *** hattwick has joined #openstack-dev | 13:34 | |
| *** blamar has quit IRC | 13:35 | |
| *** blamar has joined #openstack-dev | 13:36 | |
| *** hub_cap has joined #openstack-dev | 13:36 | |
| *** lts has joined #openstack-dev | 13:37 | |
| *** rkukura has joined #openstack-dev | 13:38 | |
| garyk | russellb: sadly it did not solve the issue. when a message is received it calls <quantum.openstack.common.rpc.amqp.ProxyCallback object at 0x2251290>. silly question - how/when is the callback stored on the proxy callback? | 13:39 |
| russellb | garyk: it is done by create_consumer | 13:40 |
| russellb | can you do some digging / add some logging to see if the code gets to where it passes off to the thread pool, and then see if that code ever runs? | 13:40 |
| garyk | russellb: ok, i'll try | 13:40 |
| *** yazirian has quit IRC | 13:45 | |
| zykes- | Daviey: Is quantum-agent-openvswitch missing dependencies of quantum-rootwrap and sudoers ? | 13:48 |
| Daviey | zykes-: quite possibly.. :/ | 13:49 |
| *** yazirian has joined #openstack-dev | 13:49 | |
| garyk | russellb: thanks! you have solved the problems. 1. my bad regarding the eventlet monkey (when I added this I saw the callback was being called) 2. the logging helped - saw that the function was not called with the correct parameters. when you have some time i would like to pick your brain regarding the openstack common and quantum. | 13:49 |
| *** yazirian has quit IRC | 13:50 | |
| *** edygarcia has joined #openstack-dev | 13:50 | |
| russellb | garyk: great! glad it's working! :) | 13:51 |
| zykes- | Daviey: where do I file a bug ? ) | 13:51 |
| russellb | garyk: pick away, now is as good of a time as any | 13:51 |
| * koolhead17 points zykes- to launchpad | 13:51 | |
| zykes- | quantum-rootwrap - should be packages with quantum-common | 13:51 |
| * koolhead17 kicks zykes- at same time 4 asking | 13:51 | |
| zykes- | and also common should have sudoers.d, agents should have upstart jobs | 13:52 |
| *** yazirian has joined #openstack-dev | 13:52 | |
| zykes- | Should it be filed under Quantum or Ubuntu ? | 13:52 |
| garyk | russellb: is there any document or reference on making the use of openstack common code. for the RPC it would be great to use this in quantum instead of duplicating every thing. i did have one or 2 minor problems with the usage of the RPC code - more specifically - there were some flags that were defined in the nova configuration and not in the RPC configuration. | 13:53 |
| koolhead17 | zykes-, are you testing it on ubuntu? | 13:54 |
| russellb | garyk: rpc isn't in openstack-common yet | 13:54 |
| garyk | russellb: i know :) | 13:54 |
| zykes- | koolhead17: yeh, gave up on Centos for now ;) | 13:54 |
| zykes- | but on ubuntu you face new issues :p | 13:54 |
| russellb | garyk: http://wiki.openstack.org/CommonLibrary | 13:54 |
| * koolhead17 gives evil laugh | 13:54 | |
| russellb | that's the doc | 13:54 |
| Daviey | zykes-: run, "ubuntu-bug quantum" | 13:54 |
| garyk | russellb: i'll be happy to help and/test | 13:55 |
| russellb | garyk: k, i also just pushed a patch series to nova for review that will change how you use rpc a bit ... https://review.openstack.org/7386 through 7391 | 13:56 |
| zykes- | garyk: can't you get a command like that for fedora / rh as well ? Would be so nice | 13:56 |
| russellb | doesn't change it conceptually | 13:56 |
| garyk | russellb: tx | 13:57 |
| garyk | zykes: not sure i understand | 13:57 |
| zykes- | garyk: the ubuntu-bug common is ok to have to collect info :) | 13:58 |
| garyk | ok | 14:00 |
| *** markmcclain has joined #openstack-dev | 14:01 | |
| zykes- | Daviey: do you know why there's: quantum-plugin-openvswitch quantum-plugin-openvswitch-agent ? | 14:01 |
| *** yazirian has quit IRC | 14:02 | |
| *** yazirian has joined #openstack-dev | 14:03 | |
| Daviey | zykes-: I do not.. zul, might be more familiar. | 14:07 |
| zul | what am i familar with? | 14:07 |
| zykes- | Seems I'm hitting https://bugs.launchpad.net/ubuntu/+source/quantum/+bug/979190 as well as missing sudoers / roowrap stuff | 14:08 |
| uvirtbot | Launchpad bug 979190 in quantum "Missing upstart scripts" [High,Fix released] | 14:08 |
| *** Mkenneth has quit IRC | 14:09 | |
| *** dtroyer is now known as dtroyer_zzz | 14:09 | |
| zul | zykes-: yeah that should be fixed in an sru | 14:10 |
| Daviey | zul: can you jump on that soonly? :) | 14:11 |
| *** mnaser has quit IRC | 14:11 | |
| zul | Daviey: yeah its on the plan for this week | 14:11 |
| *** mnaser has joined #openstack-dev | 14:11 | |
| Daviey | zul: superb | 14:11 |
| zykes- | zul: I deployed it like a few days ago to ubuntu - didn't seem to be fixed. | 14:12 |
| zykes- | SRU ? | 14:12 |
| zul | yes....it will be done in an SRU | 14:12 |
| zykes- | ok, so until the remedy is to fix stuff manually ? | 14:13 |
| zul | zykes-: yes | 14:13 |
| zykes- | does anyone have some presentational material for openstack as a whole I can use as an example ? | 14:13 |
| *** lts has quit IRC | 14:14 | |
| koolhead17 | zykes-, umm. probably slideshare can help you else in case you need any help you can poke me | 14:15 |
| zykes- | done quantum yet koolhead17 o? | 14:17 |
| koolhead17 | zykes-, http://www.slideshare.net/search/slideshow?searchfrom=header&q=openstack+essex | 14:17 |
| koolhead17 | zykes-, from coming week all on quantum :D | 14:17 |
| koolhead17 | zul, how are you now sir? | 14:17 |
| *** berendt has quit IRC | 14:18 | |
| *** mestery has joined #openstack-dev | 14:18 | |
| zul | koolhead17: jetlagged you? | 14:18 |
| koolhead17 | zul, same here. but i was more concerned about the karaoke stunt :) | 14:18 |
| zul | oh its fine | 14:18 |
| koolhead17 | aah glad to know. :) | 14:19 |
| *** dtroyer_zzz is now known as dtroyer | 14:23 | |
| *** lts has joined #openstack-dev | 14:23 | |
| *** Mkenneth has joined #openstack-dev | 14:26 | |
| *** eglynn has joined #openstack-dev | 14:27 | |
| *** shang has quit IRC | 14:28 | |
| *** eglynn__ has quit IRC | 14:29 | |
| *** eglynn_ has joined #openstack-dev | 14:31 | |
| mnaser | Can someone please review this and let me know (first non-bug-fix feature change) if there's anything to do on it? https://review.openstack.org/#/c/7376/ | 14:32 |
| bcwaldon | eglynn: updated this for you: https://review.openstack.org/#/c/7251/ | 14:32 |
| *** berendt has joined #openstack-dev | 14:33 | |
| *** mindpixel has quit IRC | 14:33 | |
| ttx | notmyname: around ? | 14:33 |
| *** berendt has quit IRC | 14:33 | |
| *** eglynn has quit IRC | 14:34 | |
| *** lloydde has joined #openstack-dev | 14:34 | |
| *** lloydde has quit IRC | 14:35 | |
| *** shang has joined #openstack-dev | 14:35 | |
| eglynn_ | bcwaldon: cool, thanks ... #7251 approved | 14:39 |
| bcwaldon | eglynn_: blame jaypipes | 14:39 |
| bcwaldon | eglynn_: https://review.openstack.org/#/c/7251/ | 14:39 |
| bcwaldon | eglynn_: approve it again! | 14:39 |
| bcwaldon | jaypipes: ^ | 14:40 |
| notmyname | ttx: ya | 14:40 |
| ttx | notmyname: there is currently a swift-bugs team which controls who can edit swift bug status (contains swift-core and the vulnerability management team). | 14:40 |
| ttx | notmyname: There is a bug subscription for swift-bugs that makes all members receive all bugmail about swift | 14:40 |
| ttx | notmyname: I'd like to seperate the right to edit Swift bugs from the obligation to receive all bugmail about Swift, so I'd like to remove swift-bugs subscription | 14:40 |
| ttx | notmyname: Currently all swift-core receives all Swift bug activity due to the swift-bugs subscription. We can preserve that by recreating the subscription at swift-core level | 14:40 |
| ttx | (if you want) | 14:40 |
| ttx | notmyname: or we can simply remove it... and let people subscribe manually if they want to. | 14:41 |
| *** crobinso has joined #openstack-dev | 14:41 | |
| ttx | notmyname: what's your preferred option ? | 14:41 |
| notmyname | ttx: I'd like swift-core to still get the emails | 14:43 |
| ttx | notmyname: ok, I'll preserve that. Will add myself temporarily to swift-core to fix that for you. | 14:44 |
| notmyname | ttx: ok | 14:44 |
| eglynn_ | bcwaldon, vishy: wrt the discussion on S3 image IDs on https://review.openstack.org/7302, what are your thoughts on explicitly exposing the image ID mapping via a new API extension? | 14:44 |
| eglynn_ | bacwaldon, vishy: e.g. https://blueprints.launchpad.net/nova/+spec/expose-s3-image-id-mapping | 14:44 |
| zykes- | garyk: do you know what part in the Quantum network manager that reserves / allocates ip's ? | 14:45 |
| eglynn_ | bcwaldon ^^^ | 14:45 |
| garyk | zykes-: at the moment quantum does not do the IPAM. this is done by melange. It is currently being merged into Quantum. | 14:45 |
| *** yazirian has quit IRC | 14:46 | |
| zykes- | garyk: I'm not using Melange, just nova_ipam | 14:46 |
| ttx | notmyname: oh. actually swift-core is already subscribed (I can see that now that I added myself to the team). So I'll just remove the swift-bugs subscription. | 14:46 |
| zykes- | but I think I'm hitting an issue where it looks for an allocated ip before updating the config file for dnsmasq but there's no ip allocated but it's reserved | 14:46 |
| ttx | notmyname: done, and removed me. | 14:47 |
| garyk | zykes-: sorry but i am not familiar with this yet. when i played with devstack today i encountered a few issues here too. i have yet to try and resolved them. | 14:48 |
| notmyname | ttx: cool, thanks | 14:50 |
| *** Gordonz has joined #openstack-dev | 14:51 | |
| *** renier has joined #openstack-dev | 14:53 | |
| *** rnirmal has joined #openstack-dev | 14:53 | |
| *** jgriffith has joined #openstack-dev | 14:57 | |
| mnaser | Vek: re the fixed ips quota patch, what do you prefer we do in this case? I know a week ago a patch got merged for key pair quotas (https://review.openstack.org/#/c/6709/) | 14:59 |
| *** byeager has quit IRC | 15:00 | |
| *** Gordonz has quit IRC | 15:01 | |
| *** Gordonz has joined #openstack-dev | 15:02 | |
| *** lloydde has joined #openstack-dev | 15:02 | |
| *** eglynn__ has joined #openstack-dev | 15:07 | |
| *** eglynn_ has quit IRC | 15:09 | |
| *** apevec has joined #openstack-dev | 15:14 | |
| *** byeager has joined #openstack-dev | 15:16 | |
| *** zaitcev has joined #openstack-dev | 15:17 | |
| *** reidrac has quit IRC | 15:20 | |
| *** heckj has joined #openstack-dev | 15:21 | |
| *** kbringard has quit IRC | 15:24 | |
| *** shang has quit IRC | 15:26 | |
| jaypipes | bcwaldon: morning. | 15:27 |
| jaypipes | bcwaldon: approved. | 15:28 |
| *** epim has joined #openstack-dev | 15:30 | |
| eglynn__ | bcwaldon: I'm slight confused by patch set 7: https://review.openstack.org/#/c/7251/7/glance/schema.py | 15:33 |
| eglynn__ | (i.e. the removal of the intersecting custom/schema properties check) | 15:33 |
| bcwaldon | eglynn__: le sigh | 15:34 |
| bcwaldon | eglynn__: apparently I dont play well with gerrit :( | 15:34 |
| bcwaldon | eglynn__ and jaypipes: https://review.openstack.org/#/c/7251/8 | 15:36 |
| bcwaldon | I'm getting tired of this | 15:36 |
| jaypipes | bcwaldon: approved (again) | 15:37 |
| eglynn__ | bcwaldon: unleash the Jenkins ;) | 15:38 |
| jaypipes | bcwaldon: so, additionalProperties is definitely not intended to provide a *user* of the schema the ability to add properties to a schema. It is intended to allow creators of subclassed schemas to add their own properties... | 15:38 |
| bcwaldon | jaypipes: define *user* of the schema | 15:39 |
| jaypipes | bcwaldon: a user of the API... i.e. Joe Schmoe. | 15:39 |
| bcwaldon | jaypipes: ok, so this isn't allowing those users to provide new schemas | 15:39 |
| jaypipes | bcwaldon: hmm... then perhaps I am not understanding the code... | 15:40 |
| bcwaldon | jaypipes: additionalProperties can do one of three things | 15:40 |
| bcwaldon | jaypipes: or even two! | 15:40 |
| jaypipes | bcwaldon: because it looks like a "properties" list is being returned in the main image dict. | 15:40 |
| bcwaldon | jaypipes: it most certainly is not :) | 15:40 |
| bcwaldon | jaypipes: there is a properties list in the database | 15:40 |
| jaypipes | bcwaldon: ok, then. I'm fully aware of the difference between {}. False, and {<some_set_of_things>} | 15:41 |
| bcwaldon | jaypipes: and there is a properties list defined in the schemas | 15:41 |
| bcwaldon | jaypipes: the test you commented on was the controller, which the serializer follows up on | 15:41 |
| bcwaldon | jaypipes: converting the 'properties' list into top-level attributes | 15:41 |
| jaypipes | bcwaldon: hmm.. | 15:41 |
| bcwaldon | jaypipes: ...depending on how the schema is defined | 15:41 |
| bcwaldon | jaypipes: possibly limiting the list | 15:41 |
| jaypipes | bcwaldon: ok, lemme go back and look... it seemed to me that the properties list resembled the v1 properties K/V list. | 15:42 |
| bcwaldon | jaypipes: the db api does, which the v2 controller passes on to the serializer | 15:43 |
| jaypipes | bcwaldon: btw, if you go to https://review.openstack.org/#/c/7316/, do you see the dependency in red, with the word [OUTDATED]? | 15:43 |
| bcwaldon | jaypipes: I think you just missed that last bit in the serializer where it gets converted | 15:43 |
| bcwaldon | jaypipes: yep | 15:43 |
| *** lloydde has quit IRC | 15:43 | |
| jaypipes | bcwaldon: do we know what that's all about? | 15:43 |
| bcwaldon | jaypipes: I'm tired of trying to keep these chains up to date at all times | 15:43 |
| bcwaldon | jaypipes: yes, the second branch in the series depends on an older patchset of the first branch | 15:44 |
| jaypipes | bcwaldon: how to resolve? | 15:44 |
| bcwaldon | jaypipes: and jenkins wont merge it until its rebased | 15:44 |
| jaypipes | bcwaldon: ah | 15:44 |
| bcwaldon | jaypipes: if we had multi-commit reviews, we wouldnt have this problem | 15:44 |
| jaypipes | bcwaldon: ok, Glancelot, I'm going to do another pass at review for the deployer properties patchset... sorry for the confusion. | 15:45 |
| bcwaldon | jaypipes: I'm glad you'r ebringing this up, I want to make sure I'm not doing anything dumb and if I'm not, that it is communicated effectively | 15:45 |
| jaypipes | bcwaldon: I think it would be good to remove the user- specified part of the commit message :) | 15:46 |
| *** jemartin has joined #openstack-dev | 15:47 | |
| bcwaldon | jaypipes: maybe that's a miscommunication, then | 15:48 |
| bcwaldon | jaypipes: so the 'allow_arbitary_attributes' option is designed to allows users to pass in arbitrary properties, but not to change the spec in any way | 15:48 |
| bcwaldon | jaypipes: its basically enabling a form of the 'properties' from the v1 spec | 15:48 |
| jaypipes | bcwaldon: hope you like my last review on https://review.openstack.org/#/c/7321/ | 15:49 |
| bcwaldon | excellent | 15:49 |
| bcwaldon | but you're going to have to review it YET AGAIN now that I have to update branch 2 in that series | 15:49 |
| jaypipes | bcwaldon: hmm, so the arbitrary properties IS like the custom K/V pairs in the v1 API?! :( | 15:50 |
| bcwaldon | jaypipes: yes, but it doesn't define a separate container! | 15:50 |
| bcwaldon | jaypipes: top-level attributes, my friend | 15:50 |
| jaypipes | bcwaldon: ok, so what DOES it do then? | 15:50 |
| bcwaldon | jaypipes: its basically a switch for changing additionalProperties from {} to false | 15:50 |
| jaypipes | bcwaldon: so a user can change the schema then! That's what I've been saying is a bad thing... | 15:50 |
| bcwaldon | jaypipes: no! | 15:50 |
| jaypipes | ugh... | 15:51 |
| jaypipes | bcwaldon: phone call? | 15:51 |
| bcwaldon | jaypipes: fine! | 15:51 |
| jaypipes | :) | 15:51 |
| *** davidha has quit IRC | 15:52 | |
| *** lts has quit IRC | 15:53 | |
| *** jemartin has left #openstack-dev | 15:54 | |
| notmyname | chmouel: I don't think setting the replica count to 1 for devstack is the best way to solve the problem | 15:54 |
| *** jakedahn_zz is now known as jakedahn | 15:54 | |
| chmouel | notmyname: the high CPU on replication problem? | 15:55 |
| notmyname | chmouel: ya. why not set it with only one device to start with? | 15:58 |
| andrewbogott_ | dtroyer: When I launch 'openstack' I get TypeError: initialize_app() takes exactly 2 arguments (1 given) | 16:00 |
| andrewbogott_ | Would you guess that that's because /I/ broke something, or because it's broken in git? | 16:00 |
| *** issackelly has joined #openstack-dev | 16:00 | |
| andrewbogott_ | (I haven't tinkered with the code, I'm just trying to add an entrypoint.) | 16:00 |
| *** lloydde has joined #openstack-dev | 16:00 | |
| chmouel | notmyname: three replicas on the same device, so using the vm_mode? | 16:01 |
| dtroyer | andrewbogott_: hmmm…not sure. what command were you trying? | 16:01 |
| notmyname | chmouel: no, 3 replicas with one device will only end up with one copy thanks to sam's recent as-unique-as-possible patch | 16:01 |
| notmyname | chmouel: I'd suggest testing that config on devstack before lowering the replica count | 16:02 |
| andrewbogott_ | I get it for everything, including a naked 'openstack' | 16:02 |
| andrewbogott_ | Maybe I have an out-of-date cliff? | 16:02 |
| chmouel | notmyname: oh I forgot about that new addition, will give it a try | 16:02 |
| dtroyer | possibly. | 16:02 |
| * andrewbogott_ doesn't know how to update cliff :( | 16:04 | |
| andrewbogott_ | Oh, wait, I guess it's just 'pip install update cliff' | 16:04 |
| andrewbogott_ | ...or not :( | 16:05 |
| *** jdurgin has joined #openstack-dev | 16:05 | |
| *** lts has joined #openstack-dev | 16:05 | |
| jgriffith | andrewbogot: Isn't it "-U" | 16:05 |
| dtroyer | -U or --upgrade | 16:06 |
| *** issackelly has quit IRC | 16:06 | |
| *** kbringard has joined #openstack-dev | 16:07 | |
| * andrewbogott_ is, for some reason, forgetting that there is such a thing as 'man' | 16:07 | |
| jgriffith | andrewbogott_: man shman | 16:07 |
| jgriffith | :) | 16:07 |
| mnaser | heckj: can you please provide a bit of insight on this? https://bugs.launchpad.net/keystone/+bug/997725 | 16:10 |
| uvirtbot | Launchpad bug 997725 in keystone "Role name is not required" [Undecided,Confirmed] | 16:10 |
| heckj | mnaser: lookiun | 16:10 |
| heckj | mnaser: your comment is dead on - little to no validation, and there should be some on all three of those areas (role names, user names, and tenant names). Additionally, user names generally need to be unique (I think that's already in there, at least with the SQL identity backend) | 16:12 |
| *** salgado is now known as salgado-lunch | 16:13 | |
| mnaser | heckj: yep, but any idea where to handle the checking so I can take care of it? I could just modify the from_dict calls or so, but that would only mean fixing it for SQL backend | 16:13 |
| mnaser | heckj: Maybe make that change in the models.py to raise an exception? Let me know if you have an idea of how you're looking to address it and I can take care of that bug | 16:14 |
| heckj | mnaser: I think the fix will need to be in each of the backends, with a general test around the driver API to verify it's working for all backends | 16:14 |
| andrewbogott_ | dtroyer: Updating cliff helped, although interactive mode seems to still be broken. | 16:15 |
| zykes- | someone here from nicira ? | 16:15 |
| dtroyer | andrewbogott_: yeah, it is for me too; I hadn't tried it in a while | 16:16 |
| mnaser | heckj: what do you think about instead adding it https://github.com/openstack/keystone/blob/master/keystone/identity/core.py -- so that the checking happens before we hit the backends? I looked aronud and in Nova it seems there is no checking for emptyness on the DB side of things, only at the API end | 16:23 |
| *** spiffxp has joined #openstack-dev | 16:23 | |
| heckj | The controllers would certainly be a reasonable place to put that code. | 16:25 |
| heckj | mnaser: ^^ | 16:25 |
| mnaser | heckj: cool, i'll look into it soon | 16:25 |
| mnaser | Also, gerrit related question.. if I want to submit a patch set that is dependent on a patchset in review, should I just git pull that patch and branch off it then when i'll submit a review, it'll set it up as a dependency? | 16:26 |
| dtroyer | andrewbogott_: Doug beat me to a fix: https://review.openstack.org/7394 | 16:26 |
| andrewbogott_ | Ah, cool. | 16:27 |
| heckj | mnaser: that'll definitely make it a dependency. Try to avoid those as much as possible (sometimes you can't) | 16:28 |
| mnaser | heckj: ok good, well I made af fix in something that was refactored so i'll make teh fix under the refactored code.. should be OK | 16:28 |
| *** openstack has joined #openstack-dev | 16:30 | |
| *** issackelly has joined #openstack-dev | 16:30 | |
| *** timello has quit IRC | 16:30 | |
| andrewbogott_ | dtroyer, dhellman_: In other news... it looks like all the existing Create implementations have a get_data method but not a run method. And yet, when I make my own Create command it seems like get_data() isn't getting called. Am I missing something? (And, is creation really supposed to happen in get_data()?) | 16:31 |
| *** maplebed has joined #openstack-dev | 16:31 | |
| *** garyk has quit IRC | 16:32 | |
| *** rbasak has quit IRC | 16:33 | |
| *** sdake has quit IRC | 16:34 | |
| *** danpb has quit IRC | 16:34 | |
| *** Chuanyong has quit IRC | 16:34 | |
| *** rods has joined #openstack-dev | 16:36 | |
| *** eglynn has joined #openstack-dev | 16:36 | |
| *** sdake has joined #openstack-dev | 16:37 | |
| *** dubsquared has joined #openstack-dev | 16:37 | |
| *** andresambrois has quit IRC | 16:37 | |
| *** eglynn__ has quit IRC | 16:37 | |
| *** davidha has joined #openstack-dev | 16:39 | |
| dtroyer | andrewbogott_: the difference is due to the mixing classes being used for displaying output. TBH I'm not clear why Doug split it up that way; I'd like to have the command patters match a little better. | 16:39 |
| bcwaldon | jaypipes and eglynn: https://review.openstack.org/#/c/7316/ can land meow | 16:39 |
| jaypipes | bcwaldon: alrready ahead of you :) | 16:40 |
| *** vincentricci has joined #openstack-dev | 16:40 | |
| bcwaldon | jaypipes: thanks a billion | 16:40 |
| jaypipes | bcwaldon: welcome a trillion. | 16:40 |
| rnirmal | jgriffith: https://review.openstack.org/#/c/7357/ | 16:41 |
| *** gael has quit IRC | 16:42 | |
| *** derekh has quit IRC | 16:44 | |
| jgriffith | rnirmal: I'll look at it now... | 16:44 |
| rnirmal | jgriffith: ok... just wanted to point it out | 16:45 |
| cp16net | jgriffith: thanks for looking at it | 16:46 |
| jgriffith | cp16net: NP | 16:46 |
| *** AlanClark has joined #openstack-dev | 16:47 | |
| *** timello has joined #openstack-dev | 16:48 | |
| andrewbogott_ | dtroyer: I still don't understand. Is it on purpose that get_data is not called for Create commands? And, if so, why is get_data implemented for all of those commands? | 16:48 |
| jaypipes | dtroyer: if you get a chance... https://review.openstack.org/#/c/7308/ :) Thanks in advance! | 16:49 |
| *** ayoung_afk has quit IRC | 16:50 | |
| vishy | eglynn_: not sure the rationale behind exposing it | 16:50 |
| *** ayoung has quit IRC | 16:50 | |
| dtroyer | jaypipes: +2 | 16:50 |
| *** ayoung has joined #openstack-dev | 16:50 | |
| jaypipes | dtroyer: thx mate | 16:51 |
| vishy | rmk: the reason for the short lease is if you use force_dhcp_release then the lease only gets dropped by dnsmasq when it times out. The default value is 2 minutes btw. | 16:51 |
| dtroyer | andrewbogott_: get_data() is called by run() from the mixing classes (Lister, Show). Because of how run() is structured, overriding it and calling the ancestor just isn't going to work, so for those commands everything has to go into get_data() | 16:52 |
| rmk | vishy: I'm not sure I follow. | 16:52 |
| vishy | blamar, jk0, pvo, comstud: do any of you know why we do a memory check on resize? | 16:53 |
| vishy | rmk: dnsmasq only expires the lease when the lease is up | 16:53 |
| andrewbogott_ | dtroyer: so you're saying I should put my code in get_data and not override run() and all, right? | 16:53 |
| vishy | rmk: so the ip address would be tied up forever | 16:53 |
| rmk | OK makes sense. | 16:53 |
| jaypipes | dolphm: https://bugs.launchpad.net/keystone/+bug/999219 | 16:53 |
| uvirtbot | Launchpad bug 999219 in keystone "Deleting a user returns IntegrityError" [Undecided,New] | 16:53 |
| vishy | rmk: if you use force_dhcp_release it gets around it by forcing a release packet | 16:53 |
| ayoung | termie, I'm looking at https://review.openstack.org/#/c/6716/ and trying to decide if it matches what the SQL driver does. | 16:54 |
| dtroyer | andrewbogott_: don't override run() if you're using get_data(). Look at the tenant commands for a working pattern... | 16:54 |
| rmk | Doesn't nova have control over the dhcp leases file anyway? Why send a packet? | 16:54 |
| andrewbogott_ | dtroyer: That's what I did first, but I put debug lines in get_data() and it was never called... | 16:54 |
| andrewbogott_ | But, anyway, I will tinker further. | 16:54 |
| vishy | rmk: so longish leases and force_dhcp_release would be fine (and recommended). Just make sure that fixed_ip_disassociate_timeout is longer than the lease time | 16:55 |
| rmk | The reason I am asking is we've had dhcp issues and I'm trying to mitigate them to some extent by increasing the lease time, while trying to understand the implications before we do. | 16:55 |
| vishy | rmk: question: are you using vlan mode? | 16:55 |
| rmk | Yes\ | 16:55 |
| *** salgado-lunch is now known as salgado | 16:55 | |
| vishy | i can explain your dhcp issues and give you a fix :) | 16:55 |
| zykes- | vishy: isn't fixed_ips supposed to be deleted if you delete a ip ? | 16:55 |
| zykes- | ehm, network | 16:55 |
| rmk | vishy: Even better :) | 16:55 |
| vishy | rmk: i do recommend you up the lease time though, the two minute default is excessive for a large deployment | 16:56 |
| rmk | Yeah it's leaving no margin for error, especially for guests which give up after one attempt. | 16:56 |
| vishy | rmk: Ok i was going to draft an email to the list about it, I was just waiting for a user to confirm that it actually fixes their problem before drafting | 16:56 |
| vishy | but i will do it now | 16:56 |
| rmk | So I suspect we're not the only ones seeing this. | 16:56 |
| vishy | rmk: the quick answer is to use the version of dnsmasq that was just released | 16:56 |
| mnaser | Can anyone give me a bit of help on how to add create a patchset that is dependent on another one? Do I just run the "git fetch" command that's listed there, do my changes and do a git review? | 16:57 |
| vishy | rmk: I will draft the email now which will explain why | 16:57 |
| rmk | 2.61 - ok | 16:57 |
| vishy | mnaser: yes, make sure your changes are in a separate commit | 16:57 |
| vishy | mnaser: and you will probably need to do git-review -R | 16:58 |
| vishy | (no-rebase) | 16:58 |
| rmk | vishy: Thanks - you're saving me a lot of research time. This has been a pretty big issue for us. :) | 16:58 |
| vishy | rmk: no worries, writing the email now | 16:58 |
| mnaser | vishy: so my workflow is, git fetch from gerrit, <fixfixfix>, git commit -a, git-review -R ? | 16:58 |
| *** anderstj has joined #openstack-dev | 16:58 | |
| vishy | mnaser: easier would bi git-review -d <review_num> | 16:59 |
| vishy | (rather than git fetch) | 16:59 |
| vishy | it pulls the review into a new branch | 16:59 |
| vishy | mnaser: otherwise yes | 16:59 |
| mnaser | vishy: I see, I was using the git fetch to get the codebase that I need from the dependent patchset, is there another way? | 17:00 |
| mtaylor | mnaser: yeah, git review -d <review_num> will do the fetch for you into a nice new branch | 17:00 |
| mtaylor | mnaser: for instance, if you want to make a change on top of https://review.openstack.org/#/c/7382/ | 17:01 |
| mtaylor | mnaser: just run "git review -d 7382, <fixfixfix, git commit -a, git review -R" | 17:01 |
| mnaser | mtaylor: sweet, that makes things easier. I will do that, thanks to vishy too | 17:02 |
| vishy | zykes-: it is after the patch went in that fixed it, although I recently saw a bug report that there is still an issue with it | 17:03 |
| rmk | vishy: I see the changelog entry in dnsmasq specific to your suggestion and Openstack. | 17:04 |
| vishy | rmk: :) | 17:10 |
| zykes- | vishy: I can confirm that :) | 17:10 |
| eglynn | vishy: sorry, missed your response a while back | 17:17 |
| eglynn | vishy: the rationale behind exposing the image mapping is to (a) enable proxies map from EC2 RunInstances to the native API call to boot the server | 17:17 |
| jgriffith | cp16net: Very minor nit pick | 17:17 |
| eglynn | vishy: & (b) allow image replication tools discover the new ami-* style ID in the destination region | 17:17 |
| vishy | eglynn: I guess I'm thinking of the ec2 component as separate (even though it is still in the core code base) | 17:18 |
| vishy | eglynn: those mapping tables belong to the ec2 layer in my mind, not core nova | 17:18 |
| eglynn | vishy: a-ha, so you're thinking the s3_images moves out of the nova DB? | 17:19 |
| vishy | eglynn: ultimately yes | 17:19 |
| eglynn | gotcha | 17:19 |
| cp16net | jgriffith: thanks i dont know why that was commented out actually.. | 17:19 |
| vishy | eglynn: I wonder if it might be better to expose the uuid as a tag or something in the ec2 layer instead | 17:19 |
| *** mcclurmc_ has quit IRC | 17:19 | |
| cp16net | jgriffith: i can fix those easily | 17:19 |
| jgriffith | cp16net: I do that all the time... just that and the size | 17:19 |
| jgriffith | cp16net: Great!! | 17:20 |
| cp16net | :) | 17:20 |
| jgriffith | cp16net: The worst is when I live a bunch of debug statements :( | 17:20 |
| jgriffith | s/live/leave/ | 17:20 |
| cp16net | heh yeah or unnecessary prints | 17:21 |
| *** markmcclain has quit IRC | 17:21 | |
| zykes- | vishy: do you know why nova-network is only logging "Using DHCP for network blah" ? | 17:21 |
| zykes- | when I create a new VM | 17:21 |
| mnaser | Vek: around? | 17:22 |
| *** harlowja has joined #openstack-dev | 17:22 | |
| *** jkyle has joined #openstack-dev | 17:24 | |
| jgriffith | cp16net: check your pep8 on the test file | 17:25 |
| *** sniperd_ has quit IRC | 17:25 | |
| *** sniperd has joined #openstack-dev | 17:25 | |
| cp16net | ok | 17:25 |
| *** Ryan_Lane has joined #openstack-dev | 17:30 | |
| *** markmcclain has joined #openstack-dev | 17:31 | |
| *** darraghb has quit IRC | 17:36 | |
| mnaser | vishy: I'm doing work on fixed IP quotas and going through the allocate_fixed_ip code, there seems to be alot of copy/pasted code between the FLAT and VLAN manager (https://github.com/openstack/nova/blob/master/nova/network/manager.py#L1187 AND https://github.com/openstack/nova/blob/master/nova/network/manager.py#L1801) --- seeing as they both are under RPCAllocateFixedIP -- do you think its ok if I create a | 17:38 |
| mnaser | _allocate_fixed_ip and put in the repeated code? | 17:38 |
| *** davidha has quit IRC | 17:38 | |
| *** davidha has joined #openstack-dev | 17:38 | |
| vishy | mnaser: sure, be careful though, there might be some differences (for example the vpn handling) | 17:40 |
| mnaser | vishy: yep, ill leave the vpn stuff alone, just the allocation part that is repeated, ill leave that in | 17:40 |
| *** andresambrois has joined #openstack-dev | 17:44 | |
| mnaser | vishy: in allocate_fixed_ip flat model, if no address is specificed, it requires an elevated context to associate ip with pool ( address = self.db.fixed_ip_associate_pool(context.elevated(), ) where as in vlan, it does not ( address = self.db.fixed_ip_associate_pool(context, ) -- is that a bug or what is it supposed to be? | 17:45 |
| mnaser | https://github.com/openstack/nova/blob/master/nova/network/manager.py#L1201 and https://github.com/openstack/nova/blob/master/nova/network/manager.py#L1817 for context | 17:46 |
| jkyle | are there any "gotchas" with the current vishy HA configuration? the last time we deployed, there were some minor bugs like missing snat rules on vm's with floating ips. | 17:47 |
| vishy | mnaser: it has to be elevated, but my guess is that it is actually elevated earlier somewhere | 17:48 |
| vishy | jkyle: the only known issue with multi_host is that you should manually disassociate and re-associate a floating ip if you are going to migrate the vm | 17:49 |
| vishy | rmk: email sent, should be showing up soon. | 17:50 |
| rmk | vishy: awesome thanks | 17:50 |
| vishy | zykes-: as opposed to what? | 17:50 |
| jkyle | vishy: awesome! good to know. | 17:50 |
| harlowja | bcwaldon: will summit another attempt at this plugabble store stuff today, haha, pain in the butt | 17:51 |
| bcwaldon | harlowja: kk | 17:51 |
| *** reed has joined #openstack-dev | 17:52 | |
| harlowja | bcwaldon: think i know why the opt was saying it was duplicated, there is a check in _is_opt_registered that does this | 17:55 |
| harlowja | if opts[opt.dest]['opt'] is not opt: | 17:55 |
| harlowja | that is doing reference equality checking, not == checking, so when u define an opt in a function, its not going to be the same opt that it was the last time that function was called | 17:55 |
| *** johnpostlethwait has joined #openstack-dev | 17:57 | |
| harlowja | makes me wonder about any opt defined in a function (since it will have the same problem) | 17:58 |
| *** johnpostlethwait has quit IRC | 17:58 | |
| harlowja | or in a local class... | 17:58 |
| *** johnpostlethwait has joined #openstack-dev | 17:58 | |
| vishy | mnaser: quick question, what is the use case for fixed ip quotas? | 17:59 |
| vishy | mnaser: are you optionally giving multiple fixed ips to each vm? | 17:59 |
| *** johnpostlethwait has quit IRC | 17:59 | |
| mnaser | vishy: hosting provider, fixed ips are public, customers pay $x per IP address, i want to limit them to x IPs total -- and yes, using multinic | 18:00 |
| mnaser | vishy: so if a customer does add-fixed-ip and ran out of ips (or tries to create a server but doesnt have enough IPs) then it would stop him | 18:00 |
| vishy | mnaser: ok, because otherwise you could just limit the instance quota | 18:01 |
| vishy | mnaser: are you using multinic with flatdhcp? | 18:01 |
| blamar | vishy: was at lunch, did you get an answer to your question about memory checks on resize? Are you talking about the scheduler? | 18:01 |
| vishy | blamar: i didn't | 18:01 |
| vishy | blamar: it is in nova/compute/api.py | 18:01 |
| blamar | checking | 18:02 |
| vishy | blamar: you were one of the people that modified the code at some point, I didn't go back and see who wrote it originally | 18:02 |
| mnaser | yes, however cant limit how much they can add extra ips, and no, multinic with flatmanager | 18:02 |
| *** GheRivero has quit IRC | 18:02 | |
| *** dubsquared has quit IRC | 18:02 | |
| *** dolphm has quit IRC | 18:03 | |
| blamar | vishy: I guess we just thought resizing to the same size instance didn't make sense? that seems to be the only checking I see at a glance | 18:04 |
| *** issackelly has quit IRC | 18:04 | |
| vishy | blamar: I agree that it doesn't make sense, but why on earth would we check ram for that | 18:04 |
| vishy | blamar: i would think we should be checking instance_type_id | 18:04 |
| vishy | blamar: it is perfectly reasonable to have two different flavors with the same ram... | 18:05 |
| blamar | vishy: yup, makes sense, I'm sold | 18:05 |
| vishy | blamar: but I wanted to make sure there wasn't some hidden xen reason for that check | 18:05 |
| *** issackelly has joined #openstack-dev | 18:05 | |
| blamar | westmaas: ^^ | 18:05 |
| *** mrunge has joined #openstack-dev | 18:07 | |
| *** dolphm has joined #openstack-dev | 18:08 | |
| *** garyk has joined #openstack-dev | 18:08 | |
| vishy | jgriffith: looks like we missed one https://bugs.launchpad.net/nova/+bug/999043 | 18:09 |
| uvirtbot | Launchpad bug 999043 in nova "Unable to retrieve details for instance due to volumes KyeError " [Undecided,New] | 18:09 |
| mnaser | vishy: sorry for q's, hope you bear with me, is there a reason in the regular network manager that we create dns entries but not in the vpn manager, missing code or it doesnt do it if using vlan manager? | 18:09 |
| mnaser | vishy: https://github.com/openstack/nova/blob/master/nova/network/manager.py#L1215 / https://github.com/openstack/nova/blob/master/nova/network/manager.py#L1801 | 18:10 |
| jgriffith | vishy: bummer... I'll grab it when I get back from lunch. | 18:10 |
| jgriffith | vishy: if it's just one I'll still be happy | 18:10 |
| *** Stackops-Jorge has quit IRC | 18:10 | |
| zykes- | vishy: what can one use for HA for instance storage ? | 18:11 |
| vishy | mnaser: I assume it was just left out by accident but andrewbogott_ may have another explanation | 18:12 |
| vishy | zykes-: SAN or ceph or sheepdog i would say | 18:12 |
| bcwaldon | eglynn: https://review.openstack.org/#/c/7316/ could use your approval | 18:12 |
| vishy | mnaser: I'm curious about your use of multi_nic. Are you doing flat or flatdhcp? | 18:14 |
| mnaser | vishy: still not deploying in production but in testing, flat + public IPs as fixed IPs, i hope that makes sense, we don't want to use NAT for floating ips/etc | 18:15 |
| mnaser | though we've considered flatdhcp, your input might be good to hear too? | 18:15 |
| bcwaldon | jaypipes and eglynn: maybe the last time we need to approve this guy: https://review.openstack.org/#/c/7321/ | 18:16 |
| vishy | mnaser: i gotcha, how are you configuring the interfaces in the guest? | 18:16 |
| zykes- | vishy: volume storage then ? | 18:17 |
| mnaser | vishy: still something that we're thinking about, however, right now we have our own proprietary platform and we have a "rebuild network" that shuts down the server, builds the appropriate network file and boots it again | 18:17 |
| vishy | zykes-: oh sorry i misread your question | 18:17 |
| mnaser | vishy: but thinking more flatdhcp might make things easier if it configures itself automatically | 18:17 |
| zykes- | vishy: I know I can use SAN yeah | 18:17 |
| zykes- | but could you have like a SAN in addition to say ISCSI for data storage ? | 18:17 |
| eglynn | bcwaldon: looking ... | 18:18 |
| bcwaldon | eglynn: thanks for sticking with me through this :/ | 18:19 |
| *** andresambrois is now known as aa | 18:20 | |
| vishy | mnaser: that sounds nasty. :) We are looking at making cloud-init set up interfaces properly | 18:21 |
| vishy | zykes-: if you just want to do HA instances, you can just back the instances dir to san / gluster / ceph /etc | 18:21 |
| mnaser | vishy: it is hell, esp on windows… I can see how cloud-init can come in very useful, and the agents too.. is cloud-init mostly for flatmanager or used for dhcp too? | 18:22 |
| zykes- | vishy: but does it do "locks" then so it doesn't do duplicates ? | 18:22 |
| hub_cap | hey does python-novaclient have a way to override the service catalog like python-glanceclient? | 18:24 |
| hub_cap | if i want to specify an endpoint and a authtoken (again, like glanceclient) | 18:24 |
| hub_cap | if not ill code it up if no one is opposed to it | 18:25 |
| hub_cap | vishy: ^ ^ | 18:28 |
| *** adjohn has joined #openstack-dev | 18:29 | |
| *** dachary has quit IRC | 18:31 | |
| *** Mkenneth has quit IRC | 18:33 | |
| *** maoy has joined #openstack-dev | 18:33 | |
| vishy | zykes-: you will need to set base_dir flag | 18:34 |
| *** alaski has quit IRC | 18:34 | |
| *** Mkenneth has joined #openstack-dev | 18:34 | |
| vishy | hub_cap: you specify a specific endpoint yes. Horizon uses it that way | 18:34 |
| rmk | vishy: which list did you send that email to? | 18:35 |
| zykes- | does that apply for sheepdogg as well ? | 18:36 |
| *** davidha has quit IRC | 18:36 | |
| vishy | zykes-: something like base_dir_name=_base_$host | 18:36 |
| vishy | zykes-: sheepdogg is for volume storage | 18:36 |
| vishy | rmk: openstack list | 18:37 |
| *** davidha has joined #openstack-dev | 18:37 | |
| rmk | You could do sheepdog with a clustered fs | 18:37 |
| rmk | for instance storage | 18:37 |
| *** alaski has joined #openstack-dev | 18:37 | |
| mnaser | rmk: isnt sheepdog block storage itself | 18:37 |
| mnaser | ? | 18:37 |
| rmk | yes | 18:37 |
| mnaser | why add a clustered fs over it to store instances on it? | 18:37 |
| rmk | distributed block storage | 18:37 |
| vishy | rmk: https://lists.launchpad.net/openstack/msg11696.html | 18:38 |
| rmk | because block storage != clustered filesystem | 18:38 |
| rmk | If you intend to make the same block device available to multiple nodes, you need a clustered filesystem | 18:38 |
| mnaser | oh, you're looking for some file system that's shared between instances? | 18:38 |
| rmk | I'm responding to zykes- :) | 18:39 |
| rmk | He's looking for HA storage solutions for instances | 18:39 |
| mnaser | oh, i see | 18:39 |
| mnaser | wasnt sure by storage = data or actaul vm drives | 18:39 |
| mnaser | anyways back to running tests :< | 18:39 |
| rmk | Yeah the term "clustered filesystem" can be misleading. | 18:40 |
| *** tryggvil_ has joined #openstack-dev | 18:41 | |
| rmk | vishy: Thanks -- odd I didn't see it, I'm on that list. | 18:41 |
| rmk | Gmail has been having problems this morning. | 18:41 |
| hub_cap | vishy: do u have an example or the parameter to do that? i dont see how to do it in the current python-novaclient | 18:42 |
| hub_cap | it seems to _always_ auth / use service catalog | 18:43 |
| vishy | hub_cap: are you talking about the command line or the library? | 18:43 |
| hub_cap | library, but also built out on the shell tool | 18:43 |
| hub_cap | but for my purposes the library | 18:43 |
| hub_cap | maybe its cuz im starting at the shell.py | 18:43 |
| *** justinsb has quit IRC | 18:44 | |
| hub_cap | even the v1_1/client.py does not have something i can pass it to specify the endpoint | 18:44 |
| rmk | zykes-: If you can go in any direction, I recommend ditching SANs and going the full commodity route with Ceph. | 18:49 |
| *** johnpostlethwait has joined #openstack-dev | 18:49 | |
| rmk | Or better yet, convince people to use root disks as ephemeral devices and use Swift + Volumes for anything which needs to persist. | 18:50 |
| *** alaski has quit IRC | 18:50 | |
| *** alaski has joined #openstack-dev | 18:50 | |
| zykes- | rmk: got a guide on Ceph? | 18:51 |
| *** justinsb has joined #openstack-dev | 18:52 | |
| rmk | zykes-: Their docs are pretty good -- http://ceph.com/docs/master/ | 18:52 |
| zykes- | docs for it with openstack ? | 18:52 |
| rmk | Nothing I'm aware of -- there isn't anything specific to openstack about this though. | 18:53 |
| *** andrewsben_zz is now known as andrewsben | 18:53 | |
| *** kbringard1 has joined #openstack-dev | 18:54 | |
| zykes- | but doesn't Ceph require images to be stored as well ? I think I read something about that | 18:57 |
| *** justinsb has quit IRC | 18:57 | |
| *** tryggvil_ has quit IRC | 18:57 | |
| hub_cap | vishy: who can i ping on how to custom craft that url and bypass the service_catalog? Id like to get down to the bottom of this :D | 18:58 |
| *** justinsb has joined #openstack-dev | 18:59 | |
| *** eglynn_ has joined #openstack-dev | 18:59 | |
| hub_cap | and i dont want to waste more of your time when i can waste someone elses! | 18:59 |
| *** dubsquared has joined #openstack-dev | 18:59 | |
| *** novas0x2a|laptop has joined #openstack-dev | 18:59 | |
| westmaas | blamar, vishy: yep agreed | 19:00 |
| *** eglynn has quit IRC | 19:01 | |
| *** edygarcia has quit IRC | 19:01 | |
| *** tserong has quit IRC | 19:02 | |
| vishy | hub_cap: look at the horizon code | 19:03 |
| hub_cap | k | 19:03 |
| *** PotHix has joined #openstack-dev | 19:03 | |
| zykes- | rmk: how you you suggest then ? | 19:03 |
| vishy | hub_cap: https://github.com/openstack/horizon/blob/master/horizon/api/nova.py#L196 | 19:04 |
| vishy | pass in auth_url directly looks like | 19:04 |
| hub_cap | sweet thx | 19:05 |
| zykes- | should one do ceph and then mount it or use the volume driver vishy ? | 19:07 |
| rmk | We really need a way to force operations via Nova. | 19:07 |
| hub_cap | Ahh so its not something they put in the actual init() code, its just magic params u have to pass thru. thx vishy | 19:07 |
| rmk | I'm finding myself directly modifying database data way too often. | 19:07 |
| *** vincentricci has quit IRC | 19:07 | |
| *** andrewsben is now known as andrewsben_zz | 19:08 | |
| rmk | Most of the disagreements between what Nova believes and actual state surrounds volumes and/or floating IPs. | 19:08 |
| *** andrewbogott_ has quit IRC | 19:08 | |
| *** torgomatic has joined #openstack-dev | 19:08 | |
| *** andrewbogott_ has joined #openstack-dev | 19:08 | |
| *** davidha has quit IRC | 19:10 | |
| *** davidha has joined #openstack-dev | 19:10 | |
| *** edygarcia has joined #openstack-dev | 19:10 | |
| *** andrewbogott_ has quit IRC | 19:11 | |
| *** andrewbogott_ has joined #openstack-dev | 19:11 | |
| *** mnewby has joined #openstack-dev | 19:11 | |
| *** joesavak has joined #openstack-dev | 19:17 | |
| *** dubsquared has quit IRC | 19:18 | |
| *** dachary has joined #openstack-dev | 19:20 | |
| davidha | Hi, does openstack support multiprocessing when running on a multi-core machien? | 19:21 |
| davidha | machine | 19:21 |
| devcamcar | davidha: can you be more specific? there are ways to accomplish that for various components | 19:23 |
| devcamcar | api services can be configured to run and make use of multiple cores | 19:24 |
| devcamcar | swift services can easily be configured to make use as well | 19:26 |
| *** ryant has joined #openstack-dev | 19:30 | |
| *** katell has joined #openstack-dev | 19:33 | |
| *** ryant has quit IRC | 19:35 | |
| *** ryant has joined #openstack-dev | 19:35 | |
| davidha | devcamcar, lets talk about Swift as an example | 19:36 |
| davidha | devcamcar, can I have multiple container servers (or account servers or object servers) serving the same directory structure? | 19:37 |
| davidha | (Not like SAIO where multiple nodes are mimiced by each server receiving its own seperate space) | 19:37 |
| davidha | In the devcamcar, in SAIO, each server has its own they would all have the same "devices" config parameter (yet different "bind_port" config parameter) | 19:39 |
| *** katell has left #openstack-dev | 19:39 | |
| *** ryant has quit IRC | 19:40 | |
| davidha | In the devcamcar, in SAIO, each server has its own "devices" param - so it is serving his own devices - this is not multiprocessing - all servers shold have the same "devices" config parameter which should include all devices on that node | 19:41 |
| harlowja | bcwaldon: https://review.openstack.org/#/c/7258/ when u get a chance | 19:41 |
| davidha | (ignore the previous sentance, it was sent as I was editing it) | 19:41 |
| bcwaldon | harlowja: I want to see what jaypipes says before I review it | 19:42 |
| harlowja | np | 19:42 |
| harlowja | jaypipes: https://review.openstack.org/#/c/7258/ | 19:42 |
| harlowja | should of fixed your comments | 19:42 |
| *** ryant has joined #openstack-dev | 19:42 | |
| devcamcar | davidha: configuring multiprocessing for SAIO seems a bit pointless :) | 19:47 |
| devcamcar | davidha: but you can choose the number of processes to run for object/proxy/container server and make use of multiple processors | 19:47 |
| annegentle | ten minutes until the Doc Team meeting in #openstack-meeting | 19:51 |
| *** spidersddd has joined #openstack-dev | 19:54 | |
| *** spidersddd has left #openstack-dev | 19:54 | |
| *** heyho has quit IRC | 19:55 | |
| *** rgoodwin has joined #openstack-dev | 19:57 | |
| davidha | devcamcar, so one can configure a second server to run on the same set of devices? Is this a config parameter or do I need to simply create a second config file and run a second process? | 20:00 |
| jaypipes | harlowja: reviewed. :) | 20:00 |
| *** Gordonz has quit IRC | 20:01 | |
| harlowja | ha, sir glancelot | 20:02 |
| harlowja | nice | 20:02 |
| devcamcar | davidha: hm, i'm not sure i fully follow what you're trying to accomplish | 20:02 |
| davidha | devcamcar: Lets say an 8 core machine with 24 disks serice as a full blown swift node (proxy + the three servers, object container and account) - Lets also assume I decided to run 8 object severs and a similar number of container servers, account servers and proxy servers - how do I accomplish that? | 20:05 |
| devcamcar | davidha: if i understand correctly, you'll want to set the "workers" option in the corresponding config files | 20:07 |
| devcamcar | davida: for example, https://github.com/openstack/swift/blob/master/etc/object-server.conf-sample#L5 | 20:07 |
| devcamcar | davidha: that will let you set the number of processes spawned by server type | 20:07 |
| devcamcar | so if you set that to 8, you'll get 8 object-server processes | 20:07 |
| davidha | devcamcar: I did not notice this param - thanks, I will have a look and trace it in the code to see what it does. This is good news though :) | 20:09 |
| *** andrewsben_zz is now known as andrewsben | 20:09 | |
| devcamcar | davidha: cool, that will get you what you want | 20:09 |
| davidha | great | 20:09 |
| zykes- | rmk: ping, should I use ceph volumes or as a mount ? | 20:09 |
| *** alaski has quit IRC | 20:14 | |
| *** alaski has joined #openstack-dev | 20:15 | |
| *** timello has quit IRC | 20:16 | |
| *** tryggvil_ has joined #openstack-dev | 20:16 | |
| *** timello has joined #openstack-dev | 20:17 | |
| *** mancdaz has quit IRC | 20:19 | |
| *** jtran has joined #openstack-dev | 20:24 | |
| koolhead17 | annegentle, around | 20:24 |
| jtran | hey all. i tried submitting a bug fix and git review complained that i didn't have an individual CLA, which I thought I had a long while ago... anyway I submitted the individual CLA, got the email replies etc and i'm still getting that error when i do git review | 20:25 |
| soren | allow_same_net_traffic unconditionally defaults to True. Am I missing something or won't that render security groups entirely useless when using flat (dhcp) networking? With flat networking, networks aren't specific for a project are they? | 20:28 |
| soren | vishy: You probably know this without even looking ^ | 20:28 |
| vishy | soren: you are correct | 20:29 |
| vishy | soren: it is leftover from when vlan was default | 20:29 |
| soren | I figured as much. | 20:29 |
| vishy | soren: (which is still the default) | 20:30 |
| vishy | soren: i would like to default to multi_host flatdhcp, since that is the most common install know afaik | 20:30 |
| vishy | * now | 20:30 |
| vishy | soren: unfortunately changing defaults is a little scary :) | 20:30 |
| * soren laughs in the face of danger | 20:31 | |
| *** stuntmachine has joined #openstack-dev | 20:32 | |
| *** vincentricci has joined #openstack-dev | 20:34 | |
| *** mrunge has quit IRC | 20:34 | |
| maoy | jtran: you need to join the openstack-cla group on Launchpad | 20:35 |
| maoy | jtran: and wait until it's manually approved | 20:36 |
| maoy | jtran: http://wiki.openstack.org/HowToContribute | 20:36 |
| *** dprince has quit IRC | 20:38 | |
| *** apevec has quit IRC | 20:43 | |
| *** openstackjenkins has quit IRC | 20:45 | |
| *** openstackjenkins has joined #openstack-dev | 20:46 | |
| *** sleepsonzzz is now known as sleepsonthefloor | 20:50 | |
| *** asalkeld_zzz is now known as asalkeld | 20:55 | |
| *** edygarcia_ has joined #openstack-dev | 20:56 | |
| *** apevec has joined #openstack-dev | 20:57 | |
| jtran | maoy, i did join that group and i think you're right i got approved by vish but i think i need one more | 20:58 |
| *** andrewsben is now known as andrewsben_zz | 20:58 | |
| vishy | jtran: most likely your emails don't match | 20:59 |
| vishy | jtran: your git email and launchpad email need to be the same | 20:59 |
| *** edygarcia has quit IRC | 20:59 | |
| *** edygarcia_ is now known as edygarcia | 20:59 | |
| jgriffith | vishy: Caught a similar miss in the ec2 version (instance foreign key still reference) | 21:01 |
| jtran | vishy, ok i'll double check that. i did change my email address lately due to company migration | 21:01 |
| jgriffith | vishy: only other thing I saw was smoketest... | 21:01 |
| vishy | jgriffith: nice | 21:01 |
| jgriffith | vishy: Tempted to just put a fixme in smoketest for now since it's all faked | 21:01 |
| jgriffith | vishy: agree/disagree? | 21:02 |
| vishy | jgriffith: smoketest is using the api | 21:02 |
| vishy | jgriffith: how does it make a difference? | 21:02 |
| jgriffith | vishy: Hmmm... | 21:02 |
| *** rkukura has quit IRC | 21:02 | |
| vishy | jgriffith: it is talking through boto, so you shouldn't have to change anything there | 21:02 |
| vishy | jgriffith: as long as ec2_api is still returning the correct values | 21:02 |
| jgriffith | vishy: Yeah, everything runs... Let me run smoketest to make sure | 21:03 |
| jgriffith | I guess that's not part of the the usual "run_tests.sh" | 21:03 |
| vishy | jgriffith: don't think you'll be able to run it easily :) | 21:03 |
| jgriffith | vishy: I was afraid of that | 21:04 |
| vishy | jgriffith: you'll have to source ec2 credentials | 21:04 |
| vishy | and then run the smoketests from the smoketests dir | 21:04 |
| cp16net | dragondm: could you review this please? https://review.openstack.org/#/c/7357/ | 21:05 |
| jgriffith | vishy: Ok, I'll try it... another "new" thing for me to learn :) | 21:05 |
| dragondm | cp16net: will do. | 21:05 |
| cp16net | thanks! | 21:05 |
| vishy | jgriffith: ok should be able to source the euca creds through devstack | 21:05 |
| vishy | jgriffith: then there should be a separate run_tests in the smoketests directory (i can't remember if there is a .sh or just a .py) | 21:06 |
| jgriffith | vishy: great! Thanks for the tips, starting up devstack now | 21:06 |
| mnaser | almost done fixed ip quotas under the new refactored quota system, woo. | 21:08 |
| *** joesavak has quit IRC | 21:10 | |
| *** justinsb has quit IRC | 21:13 | |
| *** alaski has quit IRC | 21:14 | |
| *** mdomsch has joined #openstack-dev | 21:14 | |
| *** alaski has joined #openstack-dev | 21:14 | |
| *** hub_cap has quit IRC | 21:15 | |
| *** lts has quit IRC | 21:16 | |
| *** alaski has quit IRC | 21:20 | |
| *** ywu has joined #openstack-dev | 21:21 | |
| dragondm | cp16net: change looks good. There is one confusing comment (it was copied from instance-usage, and it's wrong there too :-> ) Fix that, and I will approve. | 21:25 |
| cp16net | dragondm: should i remove it from instance-audit-usage as well in this review? | 21:26 |
| dragondm | If you like. | 21:27 |
| cp16net | dragondm: ok i updated both of the comments in instance and volume audit usage scripts | 21:29 |
| cp16net | dragondm: can you +2 it if you approve? | 21:29 |
| dragondm | ya, will, do... | 21:31 |
| dragondm | Minor nit again, tho: | 21:31 |
| cp16net | doh | 21:31 |
| cp16net | whats up? | 21:31 |
| dragondm | is should be "for volumes|instances *existing* during the audit period." | 21:32 |
| cp16net | well its a different script for each | 21:32 |
| cp16net | the same script is not for both volumes and instances | 21:32 |
| dragondm | yes, I know. I mean that the comment needs to be changes in each | 21:33 |
| dragondm | it currently says "for $item created or destroyed..." which isn't correct. | 21:33 |
| cp16net | ahh | 21:33 |
| cp16net | ok updated again. | 21:34 |
| dragondm | nit, I know, but confusing for anyone who reads the code : P | 21:34 |
| dragondm | thx | 21:34 |
| *** sleepsonthefloor is now known as sleepsonzzz | 21:34 | |
| *** dubsquared has joined #openstack-dev | 21:35 | |
| cp16net | dragondm: thanks for reviewing i thought you would be happy this is being added :) | 21:35 |
| dragondm | no prob. approved. | 21:36 |
| jgriffith | cp16net: Congrats! | 21:37 |
| cp16net | WOOT | 21:37 |
| cp16net | i will add it to that page although its on http://wiki.openstack.org/SystemUsageDataEvol | 21:37 |
| jgriffith | cp16net: Feel free to add to Cinder :) | 21:39 |
| cp16net | that should be easy enough. | 21:40 |
| jgriffith | :) | 21:40 |
| rnirmal | cp16net: use this bp https://blueprints.launchpad.net/cinder/+spec/cinder-notifications | 21:40 |
| cp16net | ok sounds good | 21:41 |
| *** crobinso has quit IRC | 21:45 | |
| *** mcclurmc_ has joined #openstack-dev | 21:46 | |
| *** stuntmachine has quit IRC | 21:47 | |
| *** maoy has left #openstack-dev | 21:47 | |
| *** apevec has quit IRC | 21:53 | |
| *** maoy has joined #openstack-dev | 21:59 | |
| *** kbringard has quit IRC | 22:02 | |
| *** mcclurmc_ has quit IRC | 22:02 | |
| andrewbogott_ | dtroyer: I'm not done with the dumb questions yet! Where should I be looking for my openstackclient log file? | 22:04 |
| dtroyer | andrewbogott_: it'll be in the current directory where you run 'openstack' | 22:05 |
| andrewbogott_ | Well, I did say it was a dumb question. | 22:05 |
| andrewbogott_ | thank you | 22:05 |
| maoy | vishy: around to talk about the check_instance_lock code? | 22:05 |
| dtroyer | andrewbogott_: NP. if you get a minute can you have a look at a couple of reviews: https://review.openstack.org/#/c/7394/ and https://review.openstack.org/#/c/7358/ | 22:06 |
| andrewbogott_ | yep, hoping to get to that as soon as get past my current blockage | 22:06 |
| dtroyer | thx | 22:06 |
| vishy | maoy: I"m here yes | 22:06 |
| maoy | vishy: great. i'm wondering what's the best behavior when there is a lock pending | 22:07 |
| *** salgado is now known as salgado-afk | 22:08 | |
| vishy | maoy: explain | 22:08 |
| maoy | vishy: if the request comes from the user, is a busy-doing-sth-else response more natural? | 22:08 |
| maoy | vishy: and some special task like terminate instance should never get the busy response | 22:09 |
| vishy | maoy: i think a busy response makes sense in most cases yes | 22:09 |
| *** cmagina_ has joined #openstack-dev | 22:10 | |
| maoy | vishy: i haven't looked at the periodic tasks to see which one (busy error vs wait) is better | 22:10 |
| comstud | We return 409s when things aren't in an appropriate state | 22:10 |
| comstud | So that might be appropriate, depending. | 22:10 |
| maoy | comstud: is that check in n-api? | 22:10 |
| comstud | Most (all?) of the current locking happens in the background on a worker. | 22:11 |
| *** jtran has quit IRC | 22:11 | |
| comstud | yea, if you raise InstanceInvalidState | 22:11 |
| comstud | the API will turn it into a 409 | 22:11 |
| comstud | If this is all worker side.. and the API doesn't need to wait for a response, it's okay to just lock.. | 22:11 |
| comstud | The API really shouldn't wait for long running tasks | 22:11 |
| comstud | they should be somehow made asynch | 22:12 |
| maoy | in theory it's possible that two API servers make conflicting decisions concurrently, right? | 22:12 |
| comstud | yes | 22:12 |
| comstud | there are currently race conditions | 22:13 |
| comstud | that we need to solve | 22:13 |
| vishy | comstud, maoy, I think we should pass all requests on to the worker and let the worker resolve them | 22:13 |
| *** cmagina has quit IRC | 22:13 | |
| *** ayoung has quit IRC | 22:13 | |
| comstud | for instance, right now API will check if instance is ina state that allows 'resize'. if it is, it'll set the state to resize. but the check and set is not atomic and it should be. | 22:14 |
| comstud | vishy, nod | 22:14 |
| vishy | comstud, maoy: api shouldn't try to figure things out because its state information may be out of date. | 22:14 |
| maoy | vishy: that will make the code simpler. | 22:14 |
| maoy | vishy: but right now the lock in worker seems to be wait by default, not fail with busy error. | 22:15 |
| comstud | vishy: agree, although that's not currently the case. | 22:15 |
| vishy | comstud: it should pass the request to the worker and the worker can decide if it is allowed and then return (starting a greenthread or some such to do the work) | 22:15 |
| comstud | :) | 22:15 |
| *** pvo-away is now known as pvo | 22:15 | |
| comstud | yeah, that's a decent solution | 22:15 |
| comstud | although that kind of becomes nasty w/ cells | 22:16 |
| maoy | how so? | 22:16 |
| vishy | comstud: the alternative is distributed locking which adds a bunch of complexity that i don't think we really need. | 22:17 |
| comstud | maybe not 'nasty'. i'd like to avoid 'call's to workers as much as possible | 22:17 |
| comstud | Ugh, no no no! | 22:17 |
| comstud | :) | 22:17 |
| comstud | I think you make the API be okay with race conditions | 22:17 |
| comstud | Let things fail later.. | 22:18 |
| comstud | it'd be nice to somehow be able to check for those failures, though. | 22:18 |
| *** Mkenneth has quit IRC | 22:20 | |
| comstud | maybe calls to cells is okay.. just need to deal with a cell potentially being unreachable at the time... | 22:20 |
| comstud | because it's on the other side of the country and someone got a little too crazy with a backhoe. | 22:20 |
| maoy | right now is there any call instead of cast from api to worker? | 22:21 |
| comstud | there are a few calls | 22:21 |
| maoy | or scheduler | 22:21 |
| comstud | the console calls come to mind | 22:22 |
| vishy | comstud: if the calls are actually async you can do that, accept the call and just update the status to failed later when the worker is reachable and says hell no | 22:22 |
| *** lifeless has quit IRC | 22:22 | |
| comstud | vishy: That's my preference long term | 22:22 |
| comstud | I think :) | 22:22 |
| vishy | comstud: although I question whether that is actually more user friendly | 22:22 |
| comstud | It's not | 22:23 |
| maoy | i see. the console calls are sync mode anyway. | 22:23 |
| vishy | comstud: does a user want a 0.1 second call and a later poll for results, or a 0.2 second call with the results? | 22:23 |
| *** anderstj has quit IRC | 22:23 | |
| comstud | vishy: Yeah, I know. :) As an API user, I'd prefer a single call for most things. However, we do have cases right now for needing some polling mechanism.. and I think those will always be there. | 22:24 |
| *** AlanClark has quit IRC | 22:24 | |
| comstud | vishy: Like a new instance creation | 22:24 |
| *** spiffxp has quit IRC | 22:24 | |
| vishy | comstud: yes, I think the same architecture works for both, it is just whether you background the call on the api side | 22:25 |
| vishy | comstud: it should still be a call on the api side so it can recover from failure, it is just whether it happens in the background or inline in the user greentrhead | 22:25 |
| comstud | using queues for calls is also somewhat problematic | 22:25 |
| comstud | since you can later get a request.. but the API connection is long gone | 22:26 |
| *** bsza has quit IRC | 22:26 | |
| comstud | but we definitely need queues for general work | 22:27 |
| comstud | anyway, kind of digressing. ;) | 22:27 |
| russellb | i wonder if amqp supports message expiration ... | 22:28 |
| comstud | i think i read something recently that it does | 22:29 |
| comstud | but I might have been dreaming | 22:29 |
| maoy | let me try to repeat what i heard: both api and worker check for VM state. API servers might have a race condition, but the workers will serialize those calls with check_instance_lock | 22:29 |
| maoy | the VM state machine runs on both API and worker I guess? | 22:30 |
| russellb | if not, we could do expiration at the application layer, assuming clocks are synchronized ... | 22:30 |
| maoy | russellb: we can always add app layer timestamps.. | 22:30 |
| russellb | yeah, that's what i mean | 22:30 |
| comstud | maoy: Mostly right.. although I'm not sure the workers really do double check the state after acquiring lock | 22:31 |
| *** tryggvil_ has quit IRC | 22:31 | |
| comstud | maoy: There's a problem right now in that API and workers share the same state data... and they really should not. | 22:31 |
| comstud | although API state data should be based on compute state data, etc | 22:32 |
| vishy | maoy: check_instance_lock does not serialize calls | 22:32 |
| comstud | ya, serialize is a bad word too | 22:32 |
| vishy | maoy: the serialization is only on 4 calls 649 @utils.synchronized(instance_uuid) | 22:32 |
| comstud | or well | 22:32 |
| vishy | maoy: and looks like that ^^ | 22:33 |
| comstud | there's no order preservation. | 22:33 |
| vishy | maoy: check_instance_lock just makes sure the instance hasn't been locked by an administrator before executing a command | 22:33 |
| maoy | vishy: got you. i misunderstood that get_lock call | 22:34 |
| *** rafaduran has joined #openstack-dev | 22:34 | |
| vishy | Vek: looks like mnaser accidentally reuploaded your changesets | 22:34 |
| vishy | Vek: when he tried to make a dependent change | 22:34 |
| russellb | to continue my aside, looks like Qpid and RabbitMQ both support setting a TTL on messages so they can be automatically removed from a queue once the TTL is hit. | 22:34 |
| maoy | vishy: so every checked_instance_lock call triggers a db call. didn't notice that.. | 22:37 |
| *** andrewsben_zz is now known as andrewsben | 22:37 | |
| *** markmc has quit IRC | 22:38 | |
| maoy | fishy, com stud: do we allow user to terminate instance no matter what state it's in right now? sorry i haven't read much of the api code.. | 22:39 |
| maoy | vishy, com stud: sorry.. damn the auto correction feature | 22:40 |
| *** cdub has quit IRC | 22:40 | |
| vishy | maoy: we don't but we should | 22:40 |
| vishy | maoy: and we should track user termination status separately from vm status | 22:41 |
| vishy | maoy: i believe it works in almost all states | 22:41 |
| maoy | vishy: why separate? | 22:42 |
| vishy | maoy: user experience, If a user terminates an instance it should be instantly gone imo | 22:43 |
| vishy | maoy: or at least instantly marked as terminated | 22:43 |
| maoy | vishy: sounds like power off vs shutdown? | 22:43 |
| vishy | maoy: power_off is different | 22:44 |
| maoy | vishy: some users might want clean shutdown in normal case | 22:44 |
| vishy | maoy: terminate means i'm done with this guest and i never want to see it again, stop billing me | 22:44 |
| *** dhellmann_ is now known as dhellmann | 22:45 | |
| maoy | vishy: got it. | 22:45 |
| maoy | vishy: i wonder if anyone is working on this. Phil Day's proposal is kind of closely related | 22:46 |
| vishy | maoy: there are lots of people working on it in different ways | 22:46 |
| *** rnirmal has quit IRC | 22:47 | |
| *** spiffxp has joined #openstack-dev | 22:47 | |
| *** pixelbeat has quit IRC | 22:47 | |
| maoy | vishy: whoever got +2 from code review wins? :) | 22:47 |
| vishy | maoy: well hopefully people are working together :) | 22:48 |
| vishy | maoy: phil's team is backlogged on their upgrade, so they won't be doing anything for a while | 22:48 |
| maoy | russellb: is there a similar feature in zeroMQ for TTL? or it's so different wthat i'm speaking nonsense. | 22:48 |
| vishy | maoy: so it is you and johannes primarily | 22:48 |
| *** pmezard has quit IRC | 22:49 | |
| maoy | vishy: got it. | 22:49 |
| vishy | maoy: zeromq uses direct connections, so ttl for a message doesn't makes sense, there is already a timeout on making the connection though | 22:49 |
| andrewbogott_ | dtroyer: What does it mean to enable or disable a tenant for a given user? | 22:50 |
| *** dubsquared has quit IRC | 22:51 | |
| dtroyer | andrewbogott_: I'm not certain of all of the effects of disabling a tenant, but it isn't user-specific. | 22:51 |
| andrewbogott_ | In https://review.openstack.org/#/c/7364/1/openstackclient/identity/v2_0/user.py you have args marked 'enable/disable tenant'. Is that a copy/paste error? | 22:52 |
| dtroyer | yup, s/tenant/user/ | 22:53 |
| maoy | vishy, comstud: thx! | 22:54 |
| mnaser | vishy: i may have fixed my little blunder, i was on the wrong branch, i switched back to the branch that i was on and doing a git diff between the last two commits only show my changes.. | 22:54 |
| *** blamar has quit IRC | 22:54 | |
| vishy | mnaser: nice | 22:55 |
| vishy | maoy: yw | 22:55 |
| mnaser | vishy: also just did a git commit ammend and not the git log has my latest one with a change id (unique to it) and git diff shows my added changes.. git-review -R is what i should try to do? | 22:56 |
| *** curryboy has joined #openstack-dev | 22:56 | |
| andrewbogott_ | dtroyer, dhellmann_: I spent the morning at the DMV, so everyone gets -1's today | 22:57 |
| *** asalkeld has quit IRC | 22:58 | |
| *** pvo is now known as pvo-away | 22:58 | |
| dtroyer | andrewbogott_: at least you're not in KS where the lines have been >6 hours after their new computer system went online last week... | 22:58 |
| *** akscram has quit IRC | 22:58 | |
| andrewbogott_ | True! | 22:58 |
| *** mdomsch has quit IRC | 22:58 | |
| *** asalkeld has joined #openstack-dev | 23:03 | |
| *** dtroyer is now known as dtroyer_zzz | 23:03 | |
| *** dubsquared has joined #openstack-dev | 23:06 | |
| *** andrewsben is now known as andrewsben_zz | 23:08 | |
| *** lloydde has quit IRC | 23:08 | |
| jgriffith | vishy: Good news is I was able to get the smoketests running... bad news is they all seem to have problems (including my instance key) | 23:10 |
| jgriffith | vishy: I'll pick through this a bit before submitting anything for the bug | 23:10 |
| *** roge has quit IRC | 23:10 | |
| *** dtroyer_zzz is now known as dtroyer | 23:13 | |
| *** gyee has joined #openstack-dev | 23:15 | |
| *** edygarcia has quit IRC | 23:26 | |
| *** salgado-afk has quit IRC | 23:30 | |
| *** lloydde has joined #openstack-dev | 23:31 | |
| *** kbringard has joined #openstack-dev | 23:45 | |
| *** lloydde has quit IRC | 23:49 | |
| *** kbringard has quit IRC | 23:51 | |
Generated by irclog2html.py 2.14.0 by Marius Gedminas - find it at mg.pov.lt!