Wednesday, 2014-06-04

*** alazarev has quit IRC00:04
*** alazarev has joined #openstack-sahara00:18
*** bradd1 has joined #openstack-sahara00:28
*** zhiyan_ is now known as zhiyan00:28
*** IlyaE has joined #openstack-sahara00:31
*** shakayumi has joined #openstack-sahara00:32
*** shakayumi has quit IRC00:45
*** ViswaV has quit IRC01:10
*** alazarev has quit IRC01:26
*** ViswaV has joined #openstack-sahara01:28
*** ViswaV_ has joined #openstack-sahara02:02
*** ViswaV has quit IRC02:05
*** IlyaE has quit IRC02:10
*** IlyaE has joined #openstack-sahara02:10
*** ViswaV_ has quit IRC02:15
*** sballe has joined #openstack-sahara03:47
*** zhiyan is now known as zhiyan_04:15
*** zhiyan_ is now known as zhiyan04:17
*** shakamunyi has joined #openstack-sahara05:04
*** bandarji has joined #openstack-sahara05:10
*** bandarji has quit IRC05:18
*** bandarji has joined #openstack-sahara05:18
openstackgerritSergey Lukjanov proposed a change to openstack/sahara: Rework keystone auth_token middleware configs  https://review.openstack.org/7468805:31
*** alazarev has joined #openstack-sahara05:36
openstackgerritOpenStack Proposal Bot proposed a change to openstack/sahara: Imported Translations from Transifex  https://review.openstack.org/9700606:01
openstackgerritSergey Kolekonov proposed a change to stackforge/sahara-ci-config: Add HDP2 job  https://review.openstack.org/9674606:04
*** bandarji has quit IRC06:35
*** alazarev has quit IRC07:01
*** skolekonov has joined #openstack-sahara07:16
openstackgerritA change was merged to openstack/sahara: Create trusts for admin user with correct tenant name  https://review.openstack.org/9332207:19
yanndegatdmitryme: no07:19
openstackgerritDmitry Mescheryakov proposed a change to openstack/python-saharaclient: Fix network passing for cluster-template-create  https://review.openstack.org/9725507:21
openstackgerritSergey Kolekonov proposed a change to stackforge/sahara-ci-config: disable Sahara pypi mirror in all jobs  https://review.openstack.org/9770007:33
*** zhiyan is now known as zhiyan_07:34
openstackgerritSergey Kolekonov proposed a change to stackforge/sahara-ci-config: Add HDP2 job  https://review.openstack.org/9674607:51
*** IlyaE has quit IRC07:55
*** jpich has joined #openstack-sahara07:55
*** julienvey has joined #openstack-sahara08:07
openstackgerritDaniele Venzano proposed a change to openstack/sahara: Documentation for the Spark plugin  https://review.openstack.org/9749808:10
openstackgerritDaniele Venzano proposed a change to openstack/sahara: Add Spark to overview and feature matrix  https://review.openstack.org/9770908:10
*** shakamunyi has quit IRC08:10
*** tosky has joined #openstack-sahara08:23
*** yanndegat has quit IRC08:30
*** julienvey has quit IRC08:35
*** IvanBerezovskiy1 has joined #openstack-sahara08:35
*** sballe has quit IRC08:48
*** openstackgerrit has quit IRC08:48
*** IvanBerezovskiy1 has quit IRC08:48
*** liusheng has quit IRC08:48
*** IvanBerezovskiy has joined #openstack-sahara08:48
*** IvanBerezovskiy has left #openstack-sahara08:49
*** sballe has joined #openstack-sahara08:49
*** IvanBerezovskiy has joined #openstack-sahara08:49
*** liusheng has joined #openstack-sahara08:49
*** openstackgerrit has joined #openstack-sahara08:50
*** julienvey has joined #openstack-sahara09:03
*** witlessb has joined #openstack-sahara09:04
*** julienvey has quit IRC09:08
*** julienvey has joined #openstack-sahara09:17
*** julienvey has quit IRC09:42
*** julienvey has joined #openstack-sahara09:43
openstackgerritSergey Lukjanov proposed a change to openstack/sahara: Rework keystone auth_token middleware configs  https://review.openstack.org/7468809:44
*** julienvey has quit IRC09:45
*** vrovachev has joined #openstack-sahara09:57
*** Timotey has joined #openstack-sahara10:02
dmitrymeyanndegat: we’ve uploaded the image to the sahara-files, here is the link: http://sahara-files.mirantis.com/centos-6_4-64-hdp-2-0.qcow210:20
*** sballe has quit IRC10:35
*** shakamunyi has joined #openstack-sahara10:37
*** sballe has joined #openstack-sahara10:40
*** shakamunyi has quit IRC10:41
*** sballe_ has joined #openstack-sahara10:42
*** sballe has quit IRC10:45
*** sballe_ has quit IRC10:48
openstackgerritSergey Lukjanov proposed a change to openstack/sahara: Rework keystone auth_token middleware configs  https://review.openstack.org/7468811:02
*** witlessb has quit IRC11:02
*** witlessb has joined #openstack-sahara11:03
*** julienvey has joined #openstack-sahara11:10
*** julienvey has quit IRC11:15
*** julienvey has joined #openstack-sahara11:17
openstackgerritDaniele Venzano proposed a change to openstack/sahara: Documentation for the Spark plugin  https://review.openstack.org/9749811:22
openstackgerritDaniele Venzano proposed a change to openstack/sahara: Add Spark to overview and feature matrix  https://review.openstack.org/9770911:22
*** shakamunyi has joined #openstack-sahara11:38
openstackgerritafazekas proposed a change to openstack/sahara: Allow HDFS data source paths without the hdfs:// scheme  https://review.openstack.org/9776211:39
*** shakamunyi has quit IRC11:43
*** yanndegat has joined #openstack-sahara11:55
dmitrymeyanndegat: we’ve uploaded the image to the sahara-files, here is the link: http://sahara-files.mirantis.com/centos-6_4-64-hdp-2-0.qcow212:24
yanndegatok12:25
yanndegatthank you12:25
*** yanndegat has left #openstack-sahara12:37
*** _elmiko is now known as elmiko13:02
*** _crobertsrh is now known as crobertsrh13:07
*** tmckay has joined #openstack-sahara13:07
*** julienvey has quit IRC13:08
*** shakamunyi has joined #openstack-sahara13:16
*** palar has joined #openstack-sahara13:27
openstackgerritDmitry Mescheryakov proposed a change to openstack/sahara: Make deleting transient clusters safe  https://review.openstack.org/9716713:38
*** julienvey has joined #openstack-sahara13:39
*** julienve_ has joined #openstack-sahara13:41
*** julienvey has quit IRC13:44
*** tellesnobrega_ has joined #openstack-sahara13:52
*** bradd1 has quit IRC14:05
*** bradd1 has joined #openstack-sahara14:06
*** skolekonov has quit IRC14:18
*** jpich has quit IRC14:50
*** sballe has joined #openstack-sahara14:50
*** sballe_ has joined #openstack-sahara14:52
*** sballe has quit IRC14:56
*** sballe__ has joined #openstack-sahara14:57
*** jpich has joined #openstack-sahara14:57
*** IlyaE has joined #openstack-sahara15:00
*** sballe_ has quit IRC15:00
*** ghenriks has quit IRC15:04
*** IlyaE has quit IRC15:05
*** sballe__ has quit IRC15:06
*** yanndegat has joined #openstack-sahara15:13
yanndegatdmitryme: hi again dmitry15:13
yanndegattried the new image15:13
yanndegati think there must be a bug somewhere15:14
yanndegatand i think it's not related to the image15:14
yanndegatwhen i deploy my cluster15:14
yanndegatmy slaves (     DATANODE15:15
yanndegat    HDFS_CLIENT15:15
yanndegat    MAPREDUCE2_CLIENT15:15
yanndegat    YARN_CLIENT15:15
yanndegat    NODEMANAGER )15:15
yanndegatmiss a /volumes/disk1/haddop/yarn directory15:15
yanndegatwe have to manually mkdir the yarn/local dir15:15
yanndegatand chown it to the yarn user15:15
yanndegathave you ever encounter this pb ?15:16
*** tellesnobrega_ has quit IRC15:17
*** ylobankov has joined #openstack-sahara15:23
dmitrymeyanndegat: nope, we didn’t. With the new image it just worked15:26
dmitrymeon the other side we used different processes set on node groups15:27
dmitrymesee here: https://review.openstack.org/#/c/95739/5/sahara/tests/integration/configs/itest.conf.sample-full15:27
dmitrymeMASTER_NODE_PROCESSES and WORKER_NODE_PROCESSES15:27
*** IvanBerezovskiy has left #openstack-sahara15:27
yanndegathow can i know if it is due to the version of sahara we use ?15:29
dmitrymeand which version do you use?15:29
dmitrymeif it is Icehouse or master, I think it is expected to just work fine15:30
yanndegatopenstack-sahara.noarch            2014.1.0-13.el6           @openstack-icehouse15:30
yanndegatpython-django-sahara.noarch        2014.1.0-3.el6            @openstack-icehouse15:30
yanndegatpython-saharaclient.noarch         0.7.0-3.el6               @openstack-icehouse15:30
dmitrymeyep, that is Icehouse15:30
yanndegatwell, the yarn local directory is not present15:31
dmitrymetry setting processes in the node groups the same way as in config above ^^15:31
dmitrymeit might help15:32
*** alazarev has joined #openstack-sahara15:33
*** ViswaV has joined #openstack-sahara15:37
yanndegatalright15:37
yanndegatwe'll try this config, but wouldn't be strange ? the only difference between our config and the one you send is15:41
yanndegaton the master : yours have theses processes GANGLIA_SERVER, NAGIOS_SERVER, OOZIE_SERVER15:42
yanndegaton the slave : yyours have these processes : ZOOKEEPER_CLIENT, PIG, OOZIE_CLIENT15:42
yanndegatthis would mean that either PIG, OOZIE or ZOOKEEPER would create the yarn directory ?15:43
openstackgerritA change was merged to openstack/python-saharaclient: removed unused methods from nova.utils  https://review.openstack.org/9593215:46
openstackgerritOpenStack Proposal Bot proposed a change to openstack/python-saharaclient: Updated from global requirements  https://review.openstack.org/9682815:52
dmitrymeyanndegat: I don’t know HDP plugin well, but I remember that when I worked with it long ago setting the right process set was essential15:57
dmitrymemaybe Hortonworks guys fixed that15:57
dmitrymebut it is worth trying15:57
yanndegati'm trying15:57
yanndegat;)15:57
yanndegathmmm16:03
yanndegatmaybe it is because16:03
yanndegati have 2 images16:03
yanndegatregistered with the tags "hdp" "2.0.6" ;)16:03
dmitrymeyanndegat: nope, it does not matter, you select just one image while launching the cluster from the UI16:05
dmitrymejust make sure you’ve picked the new one16:05
openstackgerritMatthew Farrellee proposed a change to openstack/python-saharaclient: Remove excutils from openstack-common.conf  https://review.openstack.org/9784316:09
yanndegati've picked up the good onbe16:09
yanndegatbut it is said in the docs that hdp needs absolutely the correct tags to work properly16:10
yanndegatso ...16:10
yanndegatmy cluster is still in the "starting" phase16:10
dmitrymedon’t worry about tags if cluster reached the ‘starting’ phase16:12
dmitrymecheck sahara-api log for exceptions if it is ‘starting’ for too long16:12
yanndegatalright16:13
yanndegatit is active16:13
yanndegatwait a minute....16:13
*** tellesnobrega_ has joined #openstack-sahara16:16
yanndegatok16:17
yanndegatso the yarn directory is not present16:17
yanndegateven with the same processes as listed in your conf16:17
dmitrymedo you say that cluster entered ‘active’ state?16:18
*** alazarev has quit IRC16:19
yanndegati recreated a complete cluster16:20
*** julienve_ has quit IRC16:20
dmitrymeare you sure it is actually needed?16:22
yanndegatwith the correct image and the correct processes16:22
yanndegatthe cluster is active16:22
yanndegatwhen i ssh in a slave node16:22
dmitrymes / it / the directory /16:22
yanndegatthe yarn directory is not preset16:22
yanndegatyes16:23
yanndegatbecause in the resource manager, the node is marked as unhealthy16:23
dmitrymesorry, have really small knowledge in the area16:24
yanndegatok16:24
yanndegatnevermind, we'll figure it out16:25
yanndegatthanks16:25
dmitrymeand there seems to be no guys from the Hortonworks around16:25
dmitrymeno problem16:25
*** jpich has quit IRC16:28
*** Timotey has quit IRC16:35
*** alazarev has joined #openstack-sahara16:50
*** ylobankov has quit IRC16:55
tmckaySergeyLukjanov, aignatov, ping17:07
tmckayalazarev, too17:07
alazarevtmckay: here17:08
*** tosky__ has joined #openstack-sahara17:08
*** tosky has quit IRC17:08
tmckayalazarev, hi.  Looking to talk about spark stuff, I should probably start a pad.  And we can talk about it tomorrow at the meeting.  But, here is what I have found so far ...17:09
tmckayCloudera initially implemented spark job submission in Hue using an Oozie shell action that invoked pyspark to launch a job.  Very minimal wrapper.  I found the code from late 2013 in git.17:10
tmckayBut, this means that hadoop-mapreduce is required in order to run spark jobs.  Seems bad.  It shouldn't be required.17:10
tmckaySo, by Christmas they moved to an opensource spark job server from a company called ooyala17:10
tmckayThis seems to be what everyone wants to use on the spark mail lists for a central job server17:11
tmckaySo I think the answer to our investigation of spark via Oozie is probably "No" or at the very least "Well, okay, but we're going to turn around and replace it almost immediately"17:12
*** shakamunyi has quit IRC17:13
tmckayalazarev, so this brings up a few questions.  How in openstack can we pull in an opensource job server from a non-openstack project?  If we did that, would we just be responsible for maintaining it as part of Sahara?17:13
alazarevwe can start with Oozie and think how to make edp logic engine-independant17:13
tmckaywell, that's one option17:13
tmckayIt means that the spark plugin has to be extended to install hadoop mapreduce, just to be able to run the shell action17:14
tmckayseems like a big dependency, which I don't like17:14
alazarevbecause we need at least two engines to perform such work, and facing issues with spark itself is not something we want to do during such work17:14
alazarevtmckay: agree, but this dependancy is big mostly for customers, in terms of code it is pretty easy17:16
tmckayalazarev, I almost think it would be preferable to just issue remote commands to run spark jobs from the command line, instead of going through Oozie17:16
tmckayalazarev, yes.  It might get us a quick proof of concept but I don't think it's a good long term solution.17:17
tmckayI should start the pad and we can discuss it17:17
tmckayIt concerns me that cloudera moved away from the oozie imple immediately -- October to Dec.17:17
alazarevtmckay: agree, long term we need to support spark directly17:17
tmckaythey obviously saw no future there17:18
alazarevprobably they had all infrastructure to make such move17:18
alazarevcurrently we don’t17:18
alazarevjob_manager is hardly tied on oozie17:19
tmckayalazarev, do you mean hardwired?17:19
alazarevtmckay:17:20
alazarevtmckay: yes17:20
alazarevI tried to structure the code a little bit: https://review.openstack.org/#/c/97633/17:20
tmckayalazarev, yes, I agree.  I'm wondering if it is worth looking at the refactoring first, or make spark work via oozie first and then refactor17:20
tmckayoozie is quick, maybe, but then we throw the work away17:21
alazarevwithout throwing code away it could take even more effort17:22
alazarevincremental refactoring is much better than complete rewriting of job_manager17:23
*** tosky__ is now known as tosky17:38
*** julienvey has joined #openstack-sahara17:47
*** akuznetsov has joined #openstack-sahara18:09
*** shakayumi has joined #openstack-sahara18:18
*** elmiko has quit IRC18:39
*** elmiko has joined #openstack-sahara18:42
*** akuznetsov has quit IRC18:48
*** akuznetsov has joined #openstack-sahara18:49
*** elmiko has quit IRC18:50
*** elmiko has joined #openstack-sahara18:52
*** tellesnobrega_ has quit IRC19:12
*** tellesnobrega_ has joined #openstack-sahara19:19
*** elmiko has quit IRC19:22
*** akuznetsov has quit IRC19:23
tellesnobrega_alazarev: hi, i’m working on the quotas check bug19:26
tellesnobrega_i was thinking the following, on the manager.py i would call novaclient to get the quotas and verify if there is enough based on the templates that are passed on the values dict, does that sound like a good option?19:27
alazarevtellesnobrega_: yes, that sounds good19:27
tellesnobrega_alazarev: ok19:27
*** alazarev has quit IRC19:28
SergeyLukjanovtmckay, hey, aignatov is on vacations till the end of week and I'm too tired to be useful now19:29
SergeyLukjanovbut I see non-opensource job manager - -219:29
*** elmiko has joined #openstack-sahara19:34
*** bradd1 has quit IRC19:38
*** alazarev has joined #openstack-sahara19:40
*** palar_ has joined #openstack-sahara19:40
*** palar has quit IRC19:43
*** palar_ has quit IRC19:44
*** elmiko has quit IRC19:52
openstackgerritA change was merged to openstack/sahara: [HDP] Changed test tag for HDP1 plugin  https://review.openstack.org/9748919:54
*** elmiko has joined #openstack-sahara19:58
tmckaySergeyLukjanov, well, it is *opensource*, just not *openstack* :)20:13
tmckaySergeyLukjanov, I'm going to start a pad for discussion on how we approach this.  Whether to do something quick with Oozie, and introduce hadoop-mapreduce dependency for spark clusters initially, or do something better that takes more work and skip the oozie step20:14
tmckayI lean toward "skip".20:14
tmckaybut we can talk about it tomorrow20:14
*** bandarji has joined #openstack-sahara20:27
*** tellesnobrega_ has quit IRC20:32
*** bandarji has quit IRC20:50
*** crobertsrh is now known as _crobertsrh21:01
*** bandarji has joined #openstack-sahara21:05
*** bandarji has quit IRC21:12
*** alazarev has quit IRC21:23
*** alazarev has joined #openstack-sahara21:24
*** witlessb has quit IRC21:34
*** elmiko is now known as _elmiko21:37
*** ghenriks has joined #openstack-sahara22:04
*** bradd1 has joined #openstack-sahara22:08
*** julienvey has quit IRC22:30
*** akuznetsov has joined #openstack-sahara22:31
*** tosky has quit IRC22:42
*** shakayumi has quit IRC22:55
*** julienvey has joined #openstack-sahara23:04
*** julienvey has quit IRC23:09
*** bradd1 has quit IRC23:19
*** tmckay has left #openstack-sahara23:23
*** bandarji has joined #openstack-sahara23:36
openstackgerritAndrew Lazarev proposed a change to openstack/sahara-dashboard: Fixed help text for new cluster dialog  https://review.openstack.org/9798923:54

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