15:00:45 <dmcbride> #startmeeting OPNFV Colorado Release daily 15:00:45 <collabot> Meeting started Thu Sep 15 15:00:45 2016 UTC. The chair is dmcbride. Information about MeetBot at http://wiki.debian.org/MeetBot. 15:00:45 <collabot> Useful Commands: #action #agreed #help #info #idea #link #topic. 15:00:45 <collabot> The meeting name has been set to 'opnfv_colorado_release_daily' 15:00:50 <dmcbride> #topic roll call 15:00:53 <ulik> #info Uli Kleber 15:00:57 <dmcbride> #info David McBride 15:01:10 <dali`> #info Dan Lilliehorn (ARMBand/Enea) 15:01:34 <bryan_att> #info Bryan Sullivan 15:02:55 <jmorgan1> #info Jack Morgan (Pharos) 15:03:15 <dmcbride> can someone post the link for yardstick? 15:04:40 <dmcbride> nevermind, I have it 15:05:19 <dmcbride> #topic functest and yardstick 15:05:46 <dmcbride> do we have anyone from the test team available? 15:06:53 <dmcbride> haha - I see that Jose has been playing with javascript 15:07:09 <dmcbride> we now have fancy dials instead of red and green balls 15:07:38 <dmcbride> http://testresults.opnfv.org/reporting/functest/release/colorado/index-status-apex.html 15:08:16 <dmcbride> #info functest looks good, but we still seem to be having problems with yardstick 15:08:51 <dmcbride> dali`: thanks for joining 15:08:56 <rprakash> #info rprakash 15:09:20 <dmcbride> dali`: emails I've seen from you and Bob are encouraging 15:09:29 <dmcbride> dali`: any issues you'd like to raise? 15:10:02 <dali`> Not really. Things are good. We had successful runs of both functest and yardstick in the last hour, 100% on functest and complete suite running on yardstick. 15:10:24 <dmcbride> dali`: that's great. Good news. 15:10:41 <dali`> We see spurious issues with CI from time to time (network issues, problems with floating IPs, tempest_smoke_serial issues) but all of these happen on other PODs as well. 15:12:35 <dali`> Our plan is to send you and Bob a more detailed status summary tonight, with details on scenarios etc. 15:12:48 <dmcbride> I believe that Jose will be joining us 15:13:13 <jose_lausuch> sorry I thought it was canceleted today 15:13:41 <jmorgan1> jose_lausuch: no problem, we were just talking about you ;) 15:13:52 <jose_lausuch> I hope not bad things :) 15:14:10 <jmorgan1> jose_lausuch: David likes the work you've done on javascript 15:14:28 <jose_lausuch> which work? the gauges? 15:14:33 <dmcbride> I like the dials, that's cool 15:14:56 <jose_lausuch> :) 15:15:03 <dmcbride> jose_lausuch: we still seem to be having issues with yardstick, though 15:15:11 <jose_lausuch> I think it helps to avoid confusion, before, the "red" icon was too red... 15:16:03 <dmcbride> console log indicates soft errors (SLA exceeded) 15:16:03 <dmcbride> Not clear on what yardstick is doing or if their SLAs are reasonable given our set up. 15:16:19 <dmcbride> I received this message from Greg E this morning: 15:16:30 <jose_lausuch> aha 15:16:36 <jose_lausuch> is kubi1 here? 15:16:45 <dmcbride> kubi1: ping 15:16:47 <jose_lausuch> We removed the Doctor testcase in yardstick btw 15:16:53 <jose_lausuch> so, a lot of scenarios will turn blue 15:16:55 <jose_lausuch> or green 15:17:00 <jose_lausuch> blue in jenkins 15:17:10 <jose_lausuch> can you point to the yarsdstick failures? 15:17:37 <dmcbride> unfortunately, no 15:18:00 <dmcbride> I was hoping Greg could join us to clarify, but I don't think he's online 15:18:19 <ulik> kubi is not here due to holiday in China 15:18:33 <jose_lausuch> ulik: ah ok, thanks 15:19:34 <jose_lausuch> I'm looking at http://testresults.opnfv.org/reporting/yardstick/release/colorado/index-status-fuel.html and https://build.opnfv.org/ci/view/yardstick/job/yardstick-fuel-baremetal-daily-colorado/ 15:19:52 <jose_lausuch> mixure of blue/red 15:21:09 <jose_lausuch> I can talk only about the sfc scenario with odl_l2 15:21:18 <bryan_att> dmcbride: when you get to an open spot, I have a question about https://build.opnfv.org/ci/ vs http://testresults.opnfv.org/reporting/functest/release/colorado/index-status-apex.html 15:21:18 <jose_lausuch> I think it will turn green soon 15:21:24 <jose_lausuch> but can't tell much about the others 15:21:52 <dmcbride> jose_lausuch: Greg E just joined us 15:22:23 <dmcbride> Greg_E_: jose_lausuch was asking for a pointer to the yardstick failures you mentioned in your email today 15:23:16 <dmcbride> in the mean time, bryan_att had something he wanted to bring up 15:23:20 <dmcbride> bryan_att: go ahead 15:23:35 <bryan_att> Per jenkins, the last successful Apex build was 14 days ago, yet the testresults page shows everything for functest is running fine. what is the testresults page based on? 15:23:36 <Greg_E_> https://build.opnfv.org/ci/view/fuel/job/yardstick-fuel-baremetal-daily-colorado/ 15:24:09 <bryan_att> How can functest be running fine if Apex is not working at all? Or am I misinterpreting the jenkins page? 15:24:12 <dmcbride> #link https://build.opnfv.org/ci/view/fuel/job/yardstick-fuel-baremetal-daily-colorado/ 15:24:21 <Greg_E_> nosdn-ovs, nosdn-kvm, l2-sfc 15:24:29 <Greg_E_> etc are failing 15:24:33 <jose_lausuch> Greg_E_: most of the failures I see are due to floating ips issue (not reachable) 15:24:39 <Greg_E_> issue seems to be timing 15:24:47 <jose_lausuch> odl_l2_sfc will turn blue soon, since we use Boron RC3.5 15:24:53 <dmcbride> #info ^ pointer to yardstick failures noticed by Greg_E_ 15:25:08 <Greg_E_> I thought that floats are not really supported in most scenarios 15:25:15 <bryan_att> Also, the testresults page shows sunny for copper in Apex HA scenarios - but that has been disabled and didn't work before it was. So where is the sunny indicator coming from? 15:25:33 <jose_lausuch> then the scenario owners should take responsability and decide what test cases to run 15:25:52 <jose_lausuch> if floating ips is not supported, yardstick will fail, since its based on ssh-ing a VM though a floatip 15:26:12 <jose_lausuch> bryan_att: I have an explanation for you :) 15:26:27 <jose_lausuch> but I'd like to touch 1 topic at a time 15:26:41 <bryan_att> OK, I thought it was my turn - let me know when 15:26:56 <dmcbride> sorry - that was my fault 15:27:01 <jose_lausuch> I think there are no turns 15:27:09 <bryan_att> we need a queue 15:27:16 <dmcbride> ok - lets start with Greg_E_ issue 15:27:30 <bryan_att> pretty standard IRC meeting logistics - someone has to manage the speaker/topic queue 15:27:48 <jose_lausuch> all the yardstick 'reds' that I see are showing the same error:ssh.py:256 DEBUG Ssh is still unavailable: SSHError("Exception <class 'paramiko.ssh_exception.NoValidConnectionsError'> was raised during connect. Exception value is: NoValidConnectionsError(None, 'Unable to connect to port 22 on or 10.118.101.195' 15:27:48 <dmcbride> bryan_att: yes - that would be me and I messed it up 15:27:52 <jose_lausuch> floating ip issues 15:28:08 <jose_lausuch> dmcbride: can we info that? 15:28:18 <dmcbride> everyone can info 15:28:20 <jose_lausuch> Greg_E_: do you agree? 15:28:39 <Greg_E_> if these are floating ip tests 15:28:55 <Greg_E_> what can scenario owners do to blacklist them from running 15:28:58 <dmcbride> #info jose_lausuch says: all the yardstick 'reds' that I see are showing the same error:ssh.py:256 DEBUG Ssh is still unavailable: SSHError("Exception <class 'paramiko.ssh_exception.NoValidConnectionsError'> was raised during connect. Exception value is: NoValidConnectionsError(None, 'Unable to connect to port 22 on or 10.118.101.195' 15:29:19 <jose_lausuch> Greg_E_: we did that for bgpvpn test case, where floating ips are not supported 15:29:35 <Greg_E_> do you need to do it 15:29:37 <jose_lausuch> but maybe no one is taking a look at the jenkins logs 15:29:46 <Greg_E_> or is it something that we can do ourselves 15:29:49 <dali`> As I mentioned before, we also see these issues which cause 2/3 of all functest/yardstick runs to fail for ARMBand. 15:29:53 <jose_lausuch> Greg_E_: yes, you need to specifiy which test cases to run for your scenario 15:29:57 <jose_lausuch> and change a yaml file in yardstick 15:30:03 <Greg_E_> ok 15:30:08 <bryan_att> be aware that the SSH failure may be a timing issue; ssh does not go active immediately; I have has to put delays or loops in my tests to avoid false negatives 15:30:10 <Greg_E_> I’ll let the team know 15:30:14 <dmcbride> #info Greg_E_ asks: if these are floating ip tests, what can scenario owners do to blacklist them from running 15:30:45 <bryan_att> but floating IPs work fine in Apex and JOID 15:30:59 <jose_lausuch> bryan_att: not in all the scenarios 15:31:01 <bryan_att> with SSH or any other open port 15:31:09 <jose_lausuch> I think odl_l3 also had some issues 15:31:14 <jose_lausuch> or even onos sometimes, I dont remember 15:31:35 <jose_lausuch> but I think that is the responsability of the scenario owners 15:31:41 <jose_lausuch> to have a look at and take action 15:31:47 <jose_lausuch> not the fuel/yardstick team 15:31:56 <jose_lausuch> we can just report 15:32:19 <dmcbride> jose_lausuch: so, this is a matter of scenario owners selecting the correct set of tests? 15:32:55 <jose_lausuch> dmcbride: yes and no, for the scenarios where floatip is not supported, the default yardstick test cases will fail 15:33:09 <jose_lausuch> but for those where floatips is supported, then it might be another thing 15:33:32 <dmcbride> Greg_E_: does that give you the information you need? 15:34:24 <jose_lausuch> I see for kvm scenario another type of error: error: failed to deploy stack: 'ERROR: Authentication failed: Authentication required' 15:34:29 <dmcbride> dali`: is this helpful? Have you customized the test selection in yardstick, or are you just using the defaults? 15:34:31 <jose_lausuch> so its a different issue 15:34:35 <Greg_E_> can we get somebody from yardstick team to look at the current failures and let us know the reason 15:34:44 <jose_lausuch> Greg_E_: that would be ideal 15:34:50 <Greg_E_> we are using defaults 15:35:05 <Greg_E_> nobody knows enough about yardstick to customize 15:35:11 <dali`> We use the default test selections for the scenarios. So this is kind of helpful, we won't chase that issue anymore. We just have to run many test runs, sometimes they pass. 15:35:24 <jose_lausuch> Greg_E_: I had to ask them directly to customize bgpvpn scenario tests 15:35:33 <dmcbride> dali`: that doesn't sound good 15:35:51 <Greg_E_> who can I put in touch with fuel eng team to handle this further 15:36:04 <jose_lausuch> kubi 15:36:06 <jose_lausuch> the PTL 15:36:12 <jose_lausuch> but he is OoO 15:36:31 <Greg_E_> anybody else who is working can help? 15:36:55 <Greg_E_> did he appoint a deputy while he is OOO 15:36:55 <jose_lausuch> I can help to guide you customizing your test cases 15:37:00 <Greg_E_> ok 15:37:03 <Greg_E_> that is great 15:37:10 <Greg_E_> can I have your email address 15:37:21 <jose_lausuch> jose.lausuch@ericsson.com 15:37:27 <Greg_E_> thanks 15:37:31 <jose_lausuch> np 15:37:54 <dmcbride> jorgen.w.karlsson@ericsson.com 15:37:54 <dmcbride> houjingwen@huawei.com 15:37:54 <dmcbride> wenjing_chu@dell.com 15:37:56 <dmcbride> liangqi1@huawei.com 15:37:58 <jose_lausuch> in functest we are excluding the tests with floatingips for bgpvpn scenarios and odl_l3 15:37:58 <dmcbride> jean.gaoliang@huawei.com 15:38:00 <dmcbride> vincenzo.m.riccobene@intel.com 15:38:02 <dmcbride> here's the list of committers for yardstick: 15:38:11 <jose_lausuch> jorgen is not longer working in yardstick 15:39:11 <dmcbride> ok - we are past time, but I'm happy to continue if participants are willing 15:39:22 <dmcbride> bryan_att: let's switch to your issue 15:39:36 <Greg_E_> is yardstick a mandatory for the C1 release? 15:39:49 <bryan_att> ok, just looking for clarifications as noted earlier in the channel 15:40:06 <jose_lausuch> Kubi (PTL) = jean.gaoliang@huawei.com 15:40:16 <jose_lausuch> limingjiang <limingjiang@huawei.com> is also very helpful 15:40:43 <dmcbride> Greg_E_: the philosophy for all tests for Colorado is that the test results are guidelines, but scenario owners determine whether they will release, or not 15:40:54 <Greg_E_> ok 15:41:11 <dmcbride> Greg_E_: however, scenario owners need to provide justification and explanation for failures in the release notes 15:41:24 <Greg_E_> understood 15:41:28 <dmcbride> Greg_E_: see Sofia's recent email on release notes 15:41:35 <Greg_E_> k 15:41:39 <jose_lausuch> bryan_att: answering your first question, it does not show when the scenario is broken if we do not report errors. e.G. error in healthcheck => stop CI => no error pushed to DB, reporting based on last (successfull results) 15:41:58 <jose_lausuch> that is something to be improved for sure 15:42:27 <jose_lausuch> so if a given scenario is not run during the past days, it will still count the previous runs 15:42:46 <jose_lausuch> if they are successful, then you'll have positive numbers there 15:42:48 <bryan_att> even if they were months ago? 15:42:57 <jose_lausuch> we need to show the "current" reality 15:43:08 <bryan_att> I guess you are talking about the testresults page, right? 15:43:08 <jose_lausuch> i think there is some delay or something, I'm not sure 15:43:16 <dmcbride> jose_lausuch: doesn't it use a 50-day window? 15:43:22 <bryan_att> why it shows sunny when it's not really 15:43:26 <jose_lausuch> I can ask Morgan and Serena, who implemented the dashboard 15:43:29 <dmcbride> only uses results form the past 50 days? 15:43:30 <jose_lausuch> yes 15:43:36 <jose_lausuch> I think its something like that 15:43:42 <jose_lausuch> dont remember if its 50 or whatever days 15:43:50 <bryan_att> but nearing release, the last *week* is the important timeframe 15:44:01 <jose_lausuch> bryan_att: we need to remove copper from apex, that has to be manual and we forgot 15:44:28 <bryan_att> Apex HA only 15:44:34 <jose_lausuch> but during 1 week, the scenarios are not run 4 times, there is no time to run all of them 4 times 15:44:35 <bryan_att> Non-HA is fine. 15:44:39 <jose_lausuch> ok 15:44:43 <jose_lausuch> I take a note 15:44:52 <bryan_att> ok, two weeks. but not 50 days. 15:45:04 <jose_lausuch> ok, I will propose it 15:45:12 <bryan_att> The scenarios that are planned are shown on the release wiki page 15:45:24 <bryan_att> which outlines the scenarios per project 15:45:39 <bryan_att> I took HA off that table when the issues were found 15:45:59 <bryan_att> Note also that in some HA scenarios, the Congress service is not being installed in HA mode and there, it works 15:46:10 <bryan_att> it just does not work behind HA Proxy 15:46:29 <bryan_att> So for example, in JOID, it is working in the HA scenario 15:46:45 <bryan_att> But I took it off the expected list just to there is no confusion (or less) 15:46:50 <bryan_att> that's all for me 15:47:15 <bryan_att> https://wiki.opnfv.org/display/SWREL/Colorado+scenario+inventory+and+dependencies 15:47:35 <dmcbride> ok - good discussion 15:47:47 <dmcbride> we're about 15 minutes over 15:48:08 <dmcbride> does anyone have anything else that's urgent that they would like to bring up? 15:48:31 <jose_lausuch> ok 15:48:39 <jose_lausuch> bryan_att: noted down, will talk to the team 15:48:53 <jose_lausuch> nope, I need to leave now 15:49:00 <dmcbride> ok - same time, same place on Friday 15:49:26 <dmcbride> #endmeeting