Tuesday, 2015-07-28

*** Qiming has quit IRC00:28
*** openstackgerrit has quit IRC00:46
*** openstackgerrit has joined #senlin00:47
*** Qiming has joined #senlin01:41
*** mathspanda has joined #senlin01:43
xuhaiweiQiming, morning02:03
Qiminghi, morning02:03
xuhaiweiI reported this bug, can you have a look of it? https://bugs.launchpad.net/senlin/+bug/147875302:03
openstackLaunchpad bug 1478753 in senlin "Trust can't be created" [Undecided,New]02:03
Qimingit is about your openstacksdk version02:04
Qimingyou will use the master instead of a public release, say 0.5.0 or 0.5.102:04
xuhaiweiI am using 0.5.1 now02:05
Qiminginstead of a public release, say 0.5.0 or 0.5.102:05
Qimingyou will use the master head02:06
Qimingsomething changed recently to sdk02:06
Qimingxuhaiwei, need help on switching to master head?02:09
Qiming1. cd /usr/lib/python2.7/site-packages02:10
Qiming2. rm -fr openstack02:10
Qiming3. cd /opt/stack/02:10
Qiming4. git clone git://git.openstack.org/stackforge/python-openstacksdk sdk02:11
Qiming5. cd sdk02:11
Qiming6. sudo pip install -e .02:11
xuhaiweisadly, nothing changed02:12
xuhaiwei$ pip list|grep openstacksdk openstacksdk (0.5.1.dev236, /opt/stack/python-openstacksdk)02:12
xuhaiweithis is the current version02:13
xuhaiweimy openstacksdk is installed to /usr/local/lib/python2.7/dist-packages by default02:13
xuhaiweiI usually reinstalled it by running 'python setup.py install ', but this doesn't work either02:14
Qimingwhen you do that install, you will be copying code into /usr/lib/...02:15
Qimingsince we will need to find sdk errors frequently, I'd suggest you do an install under /opt/stack/02:15
Qiming"sudo pip install -e ." will only create a "openstacksdk.egg-link" file under /usr/lib/python2.702:16
*** Zhenqi has joined #senlin02:16
Qimingthat way, you will be sure you can change sdk code and make it work if there are bugs found02:17
*** Tennyson has joined #senlin02:17
xuhaiweinow in /usr/local/lib/python2.7/disk-packages, I have openstacksdk-0.5.1.dev236-py2.7.egg-info/02:18
Qimingthat one should be deleted as well02:18
Qimingthat file was created because you did "python setup.py install"02:19
xuhaiweiI have deleted all the related files in under /usr/local/lib/python2.7/dist-packages, and reinstalled openstacksdk as you said, but it still has the same error02:23
Qimingdid you do a reinstall of senlin?02:25
Qimingjust found a problem02:26
xuhaiweiYesterday, in order to test this bug report https://launchpad.net/bugs/1478423, Yanyan asked me to delete the credential from the db02:26
openstackLaunchpad bug 1478423 in senlin " HttpException: 403 Client Error: Forbidden, At least one role should be specified" [High,Triaged]02:26
Qimingin 0.5.1 or so, the distribution name was changed from python-openstacksdk to just 'openstacksdk'02:27
xuhaiweiAfter i deleted the credentials, I got this problem02:27
Qimingokay, I'm trying02:28
Qimingokay, I'm seeing some problems, not exactly the same one as yours02:29
openstackgerritQiming Teng proposed stackforge/senlin: Fix trust listing  https://review.openstack.org/20628202:43
Qimingxuhaiwei, just fixed a problem in trust listing02:43
xuhaiweican resolve my problem?02:46
xuhaiweiQiming, the problem is solved02:52
Qimingyours is solved?02:53
xuhaiweibut there are two versions openstacksdk02:53
xuhaiweiyes, the problem I reported is resolved02:53
Qimingokay, we only use the latest version02:53
Qimingconsidering that they will do a release soon02:54
xuhaiweihow does senlin know which openstacksdk to use?02:54
xuhaiweiI think senlin is using the one installed under /usr/local/lib/python2.7/dist-packages02:56
Qimingwith that openstacksdk 1.0 is release, we can propose that to the global-requirements02:58
Qimingnow the requirements.txt file is not good02:58
Qimingit says "python-openstacksdk>=0.4.1"02:58
Qimingif there is no such a package installed, senlin will ask for one02:58
xuhaiweiI am a little confused about this, in that case, senlin can't use the latest version I think02:59
Qimingwe should replace that line with openstacksdk>=0.5.002:59
Qimingbut that will introduce a conflict with the global requirements02:59
Qimingget the points?02:59
Qimingbecause you did not follow the steps I show you02:59
xuhaiweiyour explanation above makes sense to me03:00
Qimingthis is an uneasy period for both projects03:01
xuhaiweiyes, you mean I should delete the old version?03:01
Qimingwe hope sdk will do a release soon03:01
Qimingdelete the old version03:01
Qimingor else you will be mixing code versions03:03
xuhaiweibut everytime we run senlin/pip install -e ., we will get the old version again03:03
Qiming<Qiming> now the requirements.txt file is not good03:04
Qiming<Qiming> it says "python-openstacksdk>=0.4.1"03:04
Qiming<Qiming> if there is no such a package installed, senlin will ask for one03:04
Qiming<Qiming> we should replace that line with openstacksdk>=0.5.003:04
Qimingyou won't be able to commit it03:06
Qimingbecause the gate will refuse it03:06
*** mathspanda has quit IRC03:41
*** mathspanda has joined #senlin04:18
xuhaiweiQiming, I'd like to implement the parse_sdk_exception method, you are not working on it, right?04:57
Qimingno, I'm not, xuhaiwei05:10
xuhaiweiok, should we add it in senlin/common/utils.py, or somewhere else?05:11
Qimingthere is a partial implementation there05:12
*** mathspanda has quit IRC05:14
openstackgerritQiming Teng proposed stackforge/senlin: Fix some time related function calls  https://review.openstack.org/20631806:03
lixinhuiQiming, is any brenchmark you know used by heat or someone else for scale out/in scenarios?06:19
Qiminglixinhui, no06:19
Qimingwe test it by logging into each individual servers and burn the CPU06:20
QimingI've just noticed that someone has proposed a different approach here: https://review.openstack.org/#/c/205939/1/hot/mistral/autoscaling_using_mistral.yaml06:21
Qimingit is about generating a gzip archive from urandom stream06:22
xuhaiweiit's using mistral to do auto-scaling?06:24
lixinhuiQiming, what you use to burn cpu?06:24
Qimingxuhaiwei, it is using mistral to do .... cpu burning06:25
Qimingthat is ... funny06:25
xuhaiweiinstead of ceilometer?06:26
Qimingceilometer doesn't do cpu burning06:26
xuhaiweicpu burning means consuming the cpu?06:30
Qimingpush CPU utilization to a high value so it will be deteced by ceilometer and thus notified to Heat06:30
xuhaiweijust for test?06:31
QimingI was expecting someone to show the whole process modeled as a mistral workflow06:31
Qimingthe sample template above is disappointing06:31
Zhenqinode_create has met a problem06:51
Zhenqiin nova/server.py06:52
Zhenqiaboout image.id06:52
Zhenqiand our openstacksdk is the latest06:53
QimingZhenqi, this info is not helpful for a diagnosis06:53
Qimingplease be more specific06:54
Zhenqiwrong information " AttributeError: 'NoneType' object has no attribute 'id'"06:54
Qimingany info about the line number?06:55
Zhenqiline 22006:55
Zhenqi File "/opt/stack/senlin/senlin/profiles/os/nova/server.py", line 220, in do_create06:56
openstackgerritxu-haiwei proposed stackforge/senlin: Refactor random heat stack name  https://review.openstack.org/20633206:57
xuhaiweiZhenqi, paste all the information here http://paste.openstack.org/06:57
xuhaiweigate seems very slow today06:59
Qiming-openstackstatus/#openstack-dev- NOTICE: zuul is stuck and about to undergo an emergency restart, please be patient as job results may take a long time06:59
QimingZhenqi, did you paste the information?07:04
Qimingyou will need to tell us the url where the info is pasted07:04
QimingZhenqi, you may have specified an image you don't have07:08
openstackgerritQiming Teng proposed stackforge/senlin: Fix image finding call  https://review.openstack.org/20633607:08
Qimingthere are things Senlin can improve, see the patch above ^07:08
Qiminghowever, the root cause of the error you encountered is about image not found07:09
xuhaiweithe exception handling is not good here :)07:10
Qimingya, need to improve it07:10
Qimingif the image is not found, we were expecting SDK to raise a 40407:10
QimingSDK, however, is not doing that, if we don't specify ignore_missing=False07:11
xuhaiweisdk just returns None?07:11
Qimingseems so07:11
Qimingand the 'ignore_missing' parameter was just added recently07:11
xuhaiweifor us, all the ignore_missing parameters are desired to be set False07:12
ZhenqiIn fact, I found the image in nova image-list07:14
xuhaiweican you show your profile spec Zhenqi?07:15
Zhenqithe profile example use an old version cirros-0.3.2-x86_64-uec07:18
*** Zhenqi is now known as Xiangyang07:25
lixinhuiQiming, are you using cpuburn in your own tests?07:31
QimingHaven't touch that for some time07:31
Qiminga simple one would be an infinite loop doing some useless computations07:32
lixinhuithat must be elegantly written for multi-core servers07:33
lixinhuias stress test07:33
Qimingfor cpu in 1 2 3 4; do07:36
Qiming( while true; do true; done ) &07:36
Qiminga single line can be:07:37
Qimingcat /dev/urandom | gzip > /dev/null07:37
Qimingwell, if you want some native applications for stress testing, you can find many of them07:45
openstackgerritMerged stackforge/senlin: Fix test case of environment  https://review.openstack.org/20590007:50
openstackgerritMerged stackforge/senlin: Transfer roles of context as inputs to create trust  https://review.openstack.org/20587407:50
*** mathspanda has joined #senlin08:12
lixinhuiQiming, what scenarios are targeted by heat anti-affinity08:15
lixinhuido they have requirement to affinity?08:22
lixinhuior already have scenarios08:22
Qimingwho are they>08:26
lixinhuior HEAT users08:28
lixinhuiwhat scenarios are targeted by heat affinity?08:29
Qimingheat doesn't support this yet08:30
lixinhuidesign reason?08:31
Qimingnot in its scope08:35
lixinhuiit seems that sahara used anti-affinity of HEAT08:44
*** Tennyson has quit IRC09:01
*** Xiangyang has quit IRC09:11
QimingHeat provides no anti-affinity09:13
*** mathspanda has quit IRC09:18
openstackgerritMerged stackforge/senlin: Refactor random heat stack name  https://review.openstack.org/20633209:53
*** Qiming has quit IRC10:15
openstackgerritMerged stackforge/senlin: Fix image finding call  https://review.openstack.org/20633611:01
*** Qiming has joined #senlin11:07
lixinhuiQiming, there?11:09
lixinhuiQiming, sorry for the delay, on call just now.11:54
lixinhuiI saw heat code under engin/nova, and sahara template includes the realted properties11:56
lixinhuiis there any discussion in heat community on affinity and antiaffinity11:56
lixinhuihow for these people to decide not in heat's scope11:57
Qimingif you are talking about scheduler_hints12:05
Qimingthey are just properties exposed by nova12:05
Qimingheat's job is to expose those properties faithfuly to template writers12:06
Qimingheat doesn't care about what the properties mean12:06
lixinhuiagree, please share more discussion in heat about affintiy/antiaffinity12:06
Qimingheat doesn't want to get involved in any scheduling decisions12:06
Qimingthere is no such discussion in heat12:07
Qimingyou have to work with the team for a long time to understand their philosophy12:07
lixinhuiwanna follow this: lixinhui> Qiming, what scenarios are targeted by heat anti-affinity12:08
lixinhui<Qiming> HA12:08
Qimingplease do12:09
lixinhuimaybe I just misunderstand what you mean by HA, did you mean haha that time?12:09
Qimingno, high-availability12:10
Qiminganti-affinity is just a tiny component of a HA solution12:12
Qiminga full HA solution involves: 1) failure detection 2) failure notification 3) recovery12:13
Qimingthere are many factors related to the failure detection component, for example12:14
Qiminghost failure, network failure, storage failure, VM failure and application failure12:14
Qimingeach of which needs a solution for reliable, deterministic detection12:15
Qimingif you look into recovery, there are many factors there as well12:15
Qimingyou will need to make sure the failed nodes are fenced before taking recovery measures12:16
Qimingyou will need to choose from many alternatives (maybe in sequence) for recovery12:16
Qimingfor a VM, the recovery action could be: reboot, rebuild, evacuate, migrate, recreate12:17
lixinhuiThanks for sharing, Qiming12:20
QimingI have spent about one year on high-availability12:20
lixinhuihow anti-affinity will be involved in the loop?12:20
Qiminga lot to share, ;)12:20
lixinhuiI believe :)12:21
Qiminganti-affinity is just a preparation step for the loop12:21
lixinhuiI see, you mean diffreciate the group need HA and not12:21
Qiminga group would be better placed using a policy like anti-affinity12:22
Qimingso that the service continuity can be guaranteed to a certain degree12:22
*** jruano has joined #senlin12:23
lixinhuiokay, got it12:24
QimingI have presentations on the Atlanta summit and demos on the Paris summit12:26
lixinhuido you have charts of the two talk? and vedio if record available12:27
Qimingcannot access the video12:28
Qimingbut I do have charts for both12:28
lixinhuisend them to me, xiaoming, you are a trasure12:30
lixinhuisomething wrong with 'e' on my keyboard, I must tap it heavily to input 'e'12:32
Qiminguse your toes12:32
Qimingremind me later, busy now12:32
lixinhuiokay, I will try ...12:34
*** xuhaiwei_ has joined #senlin12:39
*** Yanyan has joined #senlin12:40
*** lkarm has joined #senlin12:45
*** Yanyanhu has joined #senlin12:46
*** Yanyan has quit IRC12:48
openstackgerritQiming Teng proposed stackforge/senlin: Sample triggers using ceilometer alarm  https://review.openstack.org/20652612:49
*** Yanyan has joined #senlin12:50
*** Yanyanhu has quit IRC12:51
openstackgerritQiming Teng proposed stackforge/senlin: WIP: Trigger for ceilometer alarms  https://review.openstack.org/20652712:52
Yanyanmeeting time?13:00
lixinhuiyes, I am in13:02
openstackgerritYanyan Hu proposed stackforge/senlin: Revise some TODO items  https://review.openstack.org/20654213:42
Qiminghoho, I didn't count down to 113:56
Yanyanfinally finish the re-installation of my environment13:56
xuhaiwei_it is too hot here13:56
Yanyansufferred so much from pbr version conflict...13:57
xuhaiwei_1.3.0 should be ok i think13:57
Yanyanhave to do too much manual interference13:59
Yanyanbeijing is also very hot13:59
Yanyanjust found serveral packages are still using pbr with lower version13:59
Yanyanhave to manually handle them13:59
Yanyanand then the cliff package version can't be satisfied...13:59
xuhaiwei_anyway you solved it14:00
Yanyanyes :)14:00
xuhaiwei_i will quit ,see you guys tomorrow14:00
xuhaiwei_good night14:00
Yanyangood night14:00
Yanyanhi, Qiming, about this patch https://review.openstack.org/17035514:01
Yanyanneed we merge it? I thought it was a test code but not sure whether we need it.14:02
Qimingem, totally forget what it is about :(14:02
Qimingah, I see14:02
Qimingit is not very useful I think14:03
Qimingit was an attempt to remove request context from invocation path14:03
Qimingfeel free to abandon it14:04
Yanyanok, or we just leave it there for possible feature usage14:04
*** xuhaiwei_ has quit IRC14:04
YanyanI will also leave, talk to you guys later14:05
lixinhuiQiming, send me your decks for the two talk14:08
*** Yanyan has quit IRC14:11
openstackgerritQiming Teng proposed stackforge/senlin: WIP: Trigger for ceilometer alarms  https://review.openstack.org/20652714:52
openstackgerritMerged stackforge/senlin: Revise some TODO items  https://review.openstack.org/20654214:54
openstackgerritQiming Teng proposed stackforge/senlin: Sample triggers using ceilometer alarm  https://review.openstack.org/20652615:04
*** Qiming has quit IRC15:32
*** lkarm has quit IRC18:33
*** lkarm has joined #senlin18:34
*** jruano has quit IRC21:07
*** lkarm has quit IRC21:43
*** jruano has joined #senlin21:51
*** jruano has quit IRC22:05
*** jruano has joined #senlin22:13
*** jruano has quit IRC22:46
*** jruano has joined #senlin22:59
*** jruano has quit IRC23:09
*** Qiming has joined #senlin23:31
*** Qiming has quit IRC23:36
*** Qiming has joined #senlin23:36

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