Thursday, 2013-12-19

« Wednesday, 2013-12-18 Index Friday, 2013-12-20 »
*** IlyaE has joined #savanna00:12
*** mattf is now known as _mattf00:50
*** IlyaE has quit IRC01:15
openstackgerritMina Park proposed a change to openstack/python-savannaclient: Adding the ability to modify node group templates  https://review.openstack.org/6227601:26
openstackgerritMina Park proposed a change to openstack/python-savannaclient: Adding the ability to modify node group templates  https://review.openstack.org/6227601:34
*** nosnos has joined #savanna01:38
*** jcooley_ has quit IRC01:55
*** jcooley_ has joined #savanna02:27
*** nosnos_ has joined #savanna02:30
*** nosnos has quit IRC02:33
*** jcooley_ has quit IRC02:37
*** macjack has joined #savanna03:05
*** jcooley_ has joined #savanna04:39
*** akuznetsov has joined #savanna04:54
openstackgerritMina Park proposed a change to openstack/savanna-dashboard: Adding the ability to modify node group template  https://review.openstack.org/6194104:56
*** nosnos_ has quit IRC05:06
*** nosnos has joined #savanna05:07
openstackgerritMinkyung Maisy Kim proposed a change to openstack/savanna-dashboard: Adding the ability to modify cluster template  https://review.openstack.org/6192305:17
*** qwerty_nor has joined #savanna05:53
*** nadya has joined #savanna05:57
*** nadya is now known as Guest6233305:57
*** Guest62333 has quit IRC06:03
*** macjack has quit IRC06:16
*** qwerty_nor has quit IRC06:47
*** akuznetsov has quit IRC06:59
*** macjack has joined #savanna07:53
*** akuznetsov has joined #savanna08:22
*** macjack has quit IRC08:33
*** jcooley_ has quit IRC08:54
openstackgerritSergey Reshetnyak proposed a change to openstack/savanna: Merge volumes.py and volumes_heat.py code  https://review.openstack.org/6307609:12
*** akuznetsov has quit IRC09:26
*** macjack has joined #savanna09:32
*** dmitryme has joined #savanna09:37
*** jcooley_ has joined #savanna09:38
*** ruhe has joined #savanna09:47
*** ruhe has quit IRC09:58
*** ruhe has joined #savanna09:58
*** NikitaKonovalov has joined #savanna09:59
*** SergeyLukjanov has joined #savanna10:03
*** DinaBelova has joined #savanna10:04
*** jcooley_ has quit IRC10:09
*** jcooley_ has joined #savanna10:19
*** NikitaKonovalov has quit IRC10:28
*** NikitaKonovalov has joined #savanna10:29
*** DinaBelova has quit IRC10:30
*** DinaBelova has joined #savanna10:30
*** akuznetsov has joined #savanna10:32
*** IvanBerezovskiy has left #savanna10:46
*** jcooley_ has quit IRC10:49
*** aignatov has joined #savanna10:51
*** ruhe has quit IRC10:59
*** nosnos has quit IRC11:00
*** ruhe has joined #savanna11:04
openstackgerritVadim Rovachev proposed a change to openstack/savanna-dashboard: Renamed tox virtualenv for UI tests  https://review.openstack.org/6308511:12
*** aignatov has quit IRC11:28
*** ruhe has quit IRC11:32
*** jcooley_ has joined #savanna11:32
*** ruhe has joined #savanna11:32
*** _mattf is now known as mattf11:46
*** aignatov has joined #savanna11:47
*** IvanBerezovskiy has joined #savanna11:51
*** mattf is now known as _mattf12:00
*** dmitryme has quit IRC12:02
*** jcooley_ has quit IRC12:04
*** ruhe is now known as ruhe_12:05
*** ruhe_ has quit IRC12:06
*** venza has quit IRC12:12
*** _mattf is now known as mattf12:14
*** aignatov has quit IRC12:16
*** aignatov has joined #savanna12:20
*** ruhe has joined #savanna12:20
*** NikitaKonovalov has quit IRC12:24
*** dmitryme has joined #savanna12:29
*** venza has joined #savanna12:40
*** akuznetsov has quit IRC12:53
openstackgerritA change was merged to openstack/savanna-dashboard: Renamed tox virtualenv for UI tests  https://review.openstack.org/6308512:56
*** jcooley_ has joined #savanna12:57
*** venza has quit IRC12:58
*** NikitaKonovalov has joined #savanna13:07
*** venza has joined #savanna13:19
*** jcooley_ has quit IRC13:27
openstackgerritYaroslav Lobankov proposed a change to openstack/savanna: Integration tests related changes  https://review.openstack.org/6311213:35
*** tmckay has joined #savanna13:40
*** dmitryme_ has joined #savanna13:42
*** dmitryme has quit IRC13:44
*** dmitryme_ is now known as dmitryme13:44
*** akuznetsov has joined #savanna13:47
*** mattf is now known as _mattf13:48
*** _mattf is now known as mattf13:56
*** akuznetsov has quit IRC13:58
*** mattf is now known as _mattf14:01
*** _mattf is now known as mattf14:02
*** qwerty_nor has joined #savanna14:09
*** mattf is now known as _mattf14:14
*** jcooley_ has joined #savanna14:18
*** _mattf is now known as mattf14:34
openstackgerritAlexander Kuznetsov proposed a change to openstack/savanna: Add code to configure cluster for external hdfs  https://review.openstack.org/4782814:45
*** openstack has joined #savanna14:49
*** jcooley_ has quit IRC14:52
*** nadya_ has joined #savanna14:57
*** nadya_ has quit IRC15:04
*** DinaBelova has quit IRC15:07
*** rnirmal has joined #savanna15:08
*** DinaBelova has joined #savanna15:12
*** nadya_ has joined #savanna15:16
*** IlyaE has joined #savanna15:21
*** DinaBelova has quit IRC15:22
*** NikitaKonovalov has quit IRC15:22
*** NikitaKonovalov has joined #savanna15:30
*** DinaBelova has joined #savanna15:41
*** jcooley_ has joined #savanna15:44
*** NikitaKonovalov has quit IRC15:55
*** SergeyLukjanov has quit IRC15:56
tmckayaignatov, I just added a draft https://review.openstack.org/#/c/63156/15:59
tmckayJust the workflow generator and a test at this point, but this is what it will look like15:59
tmckayoops, forgot pep8.  I'll fix that later15:59
*** DinaBelova has quit IRC16:03
*** qwerty_nor has quit IRC16:10
*** nadya_ has quit IRC16:11
*** qwerty_nor has joined #savanna16:13
*** jcooley_ has quit IRC16:16
*** _crobertsrh is now known as crobertsrh16:22
*** ruhe has quit IRC16:28
*** aignatov has quit IRC16:28
openstackgerritA change was merged to openstack/savanna: Fixed reporting about new cluster state  https://review.openstack.org/6297116:30
*** jcooley_ has joined #savanna16:33
*** nadya_ has joined #savanna16:34
openstackgerritSergey Reshetnyak proposed a change to openstack/savanna: Extract common part of instances.py and instances_heat.py  https://review.openstack.org/6284216:35
*** SergeyLukjanov has joined #savanna16:38
*** nadya_ has quit IRC16:40
*** ruhe has joined #savanna16:47
*** aignatov has joined #savanna16:51
*** jcooley_ has quit IRC16:52
*** jcooley_ has joined #savanna16:55
*** jcooley_ has quit IRC16:58
*** SergeyLukjanov is now known as _SergeyLukjanov17:03
*** macjack has quit IRC17:04
*** _SergeyLukjanov has quit IRC17:04
*** jcooley_ has joined #savanna17:04
*** SergeyLukjanov has joined #savanna17:07
*** jcooley_ has quit IRC17:09
*** jcooley_ has joined #savanna17:15
*** dmitryme has quit IRC17:16
*** aignatov has quit IRC17:17
*** ruhe has quit IRC17:18
*** jcooley_ has quit IRC17:22
*** DinaBelova has joined #savanna17:45
*** sreshetnyak has joined #savanna17:47
*** ruhe has joined #savanna17:51
*** dmitryme has joined #savanna17:52
SergeyLukjanovteam meeting will be in #openstack-meeting-alt18:00
SergeyLukjanovin 5 minx18:00
*** aignatov has joined #savanna18:03
SergeyLukjanovsavanna guys are you around?18:04
SergeyLukjanovincorrect channel :)18:04
*** nadya has joined #savanna18:04
*** nadya is now known as Guest7488918:05
*** alazarev has joined #savanna18:05
mattftmckay, ^^18:05
*** Guest74889 is now known as nadya_18:06
tmckaythanks18:06
*** nadya_ has quit IRC18:14
*** nadya_ has joined #savanna18:15
*** DinaBelova has quit IRC18:16
*** ruhe has quit IRC18:17
*** ruhe has joined #savanna18:19
*** akuznetsov has joined #savanna18:19
*** IlyaE has quit IRC18:28
mattfalazarev, ouch re the lack of engagement on the idh plugin18:31
*** IlyaE has joined #savanna18:31
mattfsounds like you don't want to be on the hook to maintain it and expand it to do EDP18:31
mattfand i'd agree that the vendor should have that burden18:31
chris_mAlong those lines, any news on the CDH plugin?18:32
alazarevmattf: yep, exactly18:32
mattfchris_m, i've seen no movement18:33
alazarevmattf: but it will be good to have already written code merged18:33
alazarevmattf: as a start point18:34
chris_mThat's unfortunate as well. Thanks.18:34
mattfalazarev, but we shouldn't keep the code around if it is unmaintained and doesn't support features we think savanna users should expect18:35
mattfat some point it becomes just chars we carry around that no one uses18:35
alazarevIntel wants to support it, and they claim they will18:36
* mattf nods18:36
* mattf missed https://review.openstack.org/#/c/62966/ because of V-118:37
mattfalazarev, re https://review.openstack.org/#/c/62511/7/savanna/plugins/vanilla/run_scripts.py,unified18:39
mattfwhen the hadoop fs -put fails because hdfs isn't up yet, the execute_command should raise an exception, that exception will raise through oozie_share_lib. the code above will capture and retry the operation?18:40
mattf(asking saves me from walking the code to confirm myself)18:40
alazarevmattf: no, just fail18:40
alazarevmattf: but with correct message and stacktrace18:41
mattfok, so the 60-80% failure rate will result in a failed cluster start, instead of a successful but broken cluster start?18:41
alazarevmattf: yes, and https://review.openstack.org/#/c/62966/ is to fix that18:41
mattffixes by not even attempting the oozie startup until hdfs is started?18:42
mattf(at least that's what it looks like to me)18:42
alazarevmattf: exactly18:43
mattfalazarev, awesome job! making the failure obvious is great, then properly gating startup is great!18:44
mattfalazarev, should we commit both or just 62966 though?18:44
alazarevalazarev: both of them are bugs, so both need to be fixed18:45
*** IlyaE has quit IRC18:45
alazarevmattf: both of them are bugs, so both need to be fixed18:45
mattfin case the hdfs gate fails in the future, we still get a good error message from the oozie setup?18:45
mattfi'd buy that18:45
alazarevmattf: yes18:46
* mattf goes to find his +2 stamp18:46
alazarevand https://review.openstack.org/#/c/62511/ could be reverted once the issue fixed in oozie18:46
mattfalazarev, wonderful job tracking this issue down and patching it18:46
chris_mWhere's the best place to look for errors in cluster launch? I've tried both vanilla and HDP with qcow images pointed from the docs, and the status just sits at "waiting". When I attach to the images, there are no relevant processes running. After about 12 hours, the status switched to "error". If I can find the error, that will go a long way. :) I have Havana and Savanna 0.3.18:47
alazarevthe fix is already in trunk, so we need to either wait for new version or patch our oozie in imagebuilder18:47
mattfchris_m, you should check the savanna-api log. long waiting is often a network/ssh issue18:47
*** aignatov has quit IRC18:47
mattfalazarev, my preference is to wait for a new upstream release. putting things in -extras or patching as part of DIB should be a last resort18:48
chris_mmattf: /var/log/savanna/api.log on the main savanna host? Found that....thanks!18:49
mattfchris_m, that's where i put it in the RDO packaging18:49
alazarevmattf: I've filed https://bugs.launchpad.net/savanna/+bug/1262023 for that. You can leave a comment about your preference.18:50
mattfwill do18:51
openstackgerritSergey Reshetnyak proposed a change to openstack/savanna: Extract common part of instances.py and instances_heat.py  https://review.openstack.org/6284218:57
*** rnirmal has quit IRC19:02
*** aignatov has joined #savanna19:07
tmckayaignatov, hi19:09
aignatovtmckay: I'm here :)19:12
tmckayaignatov, hi.  I'm thinking about input and output data sources for the java action19:13
*** ruhe has quit IRC19:14
tmckaymap-reduce uses mapred.input.dir (and output), hive and pig use params INPUT and OUTPUT19:14
tmckayaignatov, but java actions have only <arg>.  So, for the Wordcount example, I am passing hdfs paths as arg1 and arg2 for input/output19:15
tmckayWe have this semi-hardcoded concept of input/output in Savanna, and for ephemeral clusters we have data starting and ending in swift19:15
tmckaySo I see a few possibilities:19:16
tmckay1) use a long running cluster for a java action and use hdfs for input and output19:16
aignatovmaybe we just should not append datasources to java actions? :)19:17
tmckay2) pass swift urls as args to a java main() that expects swift urls instead of (or as an alternative to) hdfs19:17
tmckay3) move data from swift to hdfs and back again (somehow)19:18
tmckayaignatov, yes, but what if someone wants to run a Wordcount type app on an ephemeral cluster?  No solution right now, except #2 above (a specially written Wordcount that can do swift)19:19
*** akuznetsov has quit IRC19:20
aignatovbtw, did you ses how jar job works with args paver swift19:20
aignatovpaver -> over19:20
aignatovhadoop distcp -D fs.swift.service.savanna.username=admin  -D fs.swift.service.savanna.password=swordfish swift://integration.savanna/temp swift://integration.savanna/temp119:21
alazarevmattf: I've uploaded patch to https://issues.apache.org/jira/browse/HADOOP-1003219:21
tmckayaignatov, I think so.  We have those extra configs in the workflows for swift paths19:21
tmckayaignatov, but I'm assuming that the magic happens because of the hadoop patches which makes mapred.input.dir work, for example.  Not sure how Wordcount would deal with a swift input path.  I think the code in main() would have to be modified19:23
mattfalazarev, thanks19:23
tmckayunless FileInputFormat.addInputPath knows how to deal with swift://19:23
tmckaythat's the part that is troubling me19:23
aignatovthe third approach will not work for big data stored in swift defenitely19:24
tmckayaignatov, agreed, so I think that #1 (long running cluster, use hdfs) and #2 (write special main()) are the options for java actions with file io19:25
tmckayand the UI/REST will need a way to allow a  user to specify a list of args (and also the main class argument)19:25
tmckayIt was unclear to me what the swift integration patches to hadoop made available to jobs19:25
tmckayIt would be great if the Java file classes could handle swift directly19:26
tmckayFor a first pass, I suppose I can use a long running cluster and hdfs files19:27
tmckayBut we should think about all this ^^ :)19:27
tmckayso I need to go and look and make sure that EDP is not going to try to force input/output for java jobs, heh19:28
aignatovyes, actually we need  get more help from experts like Nadya and akusnetsov19:29
aignatov:)19:29
tmckaynadya_, ^^19:29
tmckayyes, agreed.  I'll pursue #1 for now19:29
*** nadya_ has quit IRC19:30
tmckayshe didn't want to answer ;-)19:31
aignatovindeed...19:31
* tmckay goes to think about how to include args and main-class in a job JSON cleanly19:31
aignatovmaybe we just should try semi-hardcoded approach :)19:32
aignatovas you said before19:32
aignatovjust claim users to pass args1 and args 2 as input/output19:32
*** sreshetnyak has quit IRC19:33
aignatovso solution 2 would work :) tmckay what do you think?19:34
aignatovand if end users will want another solution to run EDP for java actions we can rework it19:36
tmckayaignatov, yes, I think that might be okay.  We give them 2 options, hdfs with long-running cluster, no changes necessary, or swift-enabled main() for ephemeral clusters (but the user must do the work)19:39
tmckayaignatov, it makes me wonder what kind of Swift libs there are in Java.  Is #2 easy, is the question.19:39
aignatovyou mean swift libs in hadoop stuff?19:40
tmckayyes.  If I have Wordcount.main() for example, how easy or hard is it to take swift file args and work on them?19:41
tmckayI think it must not be too hard because mapred can do it :)19:42
tmckayI should go look and see how that works19:42
tmckayaignatov, these would be good user dev pages too, maybe.  Show examples of how to write different types of jobs and launch them from Savanna19:43
aignatovtmckay: as I know, map reduce example of word count can work over swift, it's easy as I remember19:45
aignatovso the example job of workout should look like:19:47
aignatovhadoop jar wordcount -D fs.swift.service.savanna.username=admin  -D fs.swift.service.savanna.password=swordfish swift://integration.savanna/input swift://integration.savanna/output19:47
aignatovsomething like that19:47
tmckayoh, wow, that works?19:48
tmckayWith a wordcount like this? http://grepcode.com/file/repository.cloudera.com/content/repositories/releases/com.cloudera.hadoop/hadoop-examples/0.20.2-320/org/apache/hadoop/examples/WordCount.java19:48
tmckayThat would me that the FileInputFormat etc already deal with the swift:// schema19:49
tmckaywhich would be great19:49
tmckay"that would mean"19:49
tmckayI'll make a cluster and try it19:49
aignatovtmckay: yeah, that's should work :) sorry, I didn't understand initially your question about swift args19:54
* aignatov is starting cluster to check his words19:54
tmckayaignatov, no problem, I'm learning as I go!  Always that way on the leading edge, it seems :)  If that works, we're all set.19:55
tmckayaignatov, heh, reading back I see now that was what you were saying earlier about hadoop distcp.  Sorry I missed it.20:15
tmckaysometimes IRC is not great (but cheaper than plan tickets)20:15
tmckayplane20:16
*** dmitryme has quit IRC20:22
openstackgerritA change was merged to openstack/savanna: Changing oozie libs setup to manual copy  https://review.openstack.org/6251120:26
*** IlyaE has joined #savanna20:28
openstackgerritAlexander Ignatov proposed a change to openstack/savanna: Fixed issue with undeleted instanses  https://review.openstack.org/6320720:31
openstackgerritAlexander Ignatov proposed a change to openstack/savanna: Fixed issue with undeleted instanses  https://review.openstack.org/6320720:45
openstackgerritAlexander Ignatov proposed a change to openstack/savanna: Fixed issue with undeleted instances  https://review.openstack.org/6320720:48
*** DinaBelova has joined #savanna20:48
*** aignatov has quit IRC20:48
*** DinaBelova has quit IRC20:52
*** alazarev has quit IRC20:53
*** alazarev has joined #savanna20:54
*** dmitryme has joined #savanna21:06
*** dmitryme has quit IRC21:29
*** SergeyLukjanov has quit IRC21:46
*** crobertsrh is now known as _crobertsrh21:48
*** qwerty_nor has quit IRC21:58
*** jcooley_ has joined #savanna22:19
*** mattf is now known as _mattf22:57
« Wednesday, 2013-12-18 Index Friday, 2013-12-20 »

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