Thursday, 2018-09-06

*** spartakos has quit IRC00:30
*** rcernin has quit IRC00:50
*** rcernin has joined #openstack-lbaas00:51
*** yamamoto has joined #openstack-lbaas00:55
openstackgerritJacky Hu proposed openstack/octavia-tempest-plugin master: Fix test_with_ipv6 = False failure
*** hongbin has joined #openstack-lbaas01:51
*** sapd1_ has joined #openstack-lbaas02:03
*** sapd1 has quit IRC02:07
*** sapd1_ has quit IRC02:09
*** spartakos has joined #openstack-lbaas02:10
*** sapd1 has joined #openstack-lbaas02:13
lxkongjohnsom: could you please have a look at The member address validation is missing for graph lb creation.02:31
lxkongwhich is urgent for us02:32
johnsomlxkong Ok, thanks for letting me know it is urgent.  It's a bit busy for me right now. I will try to take a look soon02:34
lxkongjohnsom: appreciate. After this is merged, I can continue to fix the batch member update function.02:35
johnsomOk cool.  Yeah, glad to see more validations02:35
lxkongBTW, for batch member update function, i found a bug here!/story/200360802:35
johnsomYeah, that is an odd one. Are you working on that?02:41
lxkongjohnsom: not yet. Currently I am dealing with our octavia related tickets in our production02:58
lxkongand octavia support in various orchestration tooling support, gophercloud, ansible, etc.02:59
lxkongjohnsom: fyi, ansible support here, and most of octavia functionality for gophercloud has been merged.03:00
johnsomyeah, I'm still working at 8pm on a nasty bug....03:00
johnsomlxkong Excellent!03:00
johnsomDo we have that on our SDK list?03:00
lxkongjohnsom: didn't check :-(03:00
johnsomYeah, it's there03:01
lxkongi guess gophercloud is on the list, but ansible is not03:01
lxkongthere are still some modules i'm working on03:01
lxkongcurrently, the graph creation support03:01
johnsomYeah, ansible is one of the deployment tools. It will be on that list once I write the install guide....  sigh03:01
lxkongafter the basic modules are done, i will let you know03:02
*** spartakos has quit IRC03:35
*** ramishra has joined #openstack-lbaas03:39
*** hongbin has quit IRC04:01
*** yamamoto has quit IRC04:05
*** yamamoto has joined #openstack-lbaas04:05
*** yamamoto has quit IRC04:16
*** yboaron_ has joined #openstack-lbaas04:34
openstackgerritMichael Johnson proposed openstack/octavia master: Fix health manager performance regression
openstackgerritMichael Johnson proposed openstack/octavia master: Fix health manager performance regression
openstackgerritMichael Johnson proposed openstack/octavia master: Fix health manager performance regression
*** pcaruana has joined #openstack-lbaas06:25
openstackgerritReedip proposed openstack/octavia-tempest-plugin master: Add configuration support for skipping tests
*** rcernin has quit IRC07:00
*** Emine has joined #openstack-lbaas07:06
*** velizarx has joined #openstack-lbaas07:14
*** velizarx has quit IRC07:27
*** pcaruana has quit IRC07:36
*** luksky has joined #openstack-lbaas07:41
*** velizarx has joined #openstack-lbaas07:44
*** pcaruana has joined #openstack-lbaas07:48
*** velizarx has quit IRC07:58
*** ccamposr has joined #openstack-lbaas08:04
*** celebdor has joined #openstack-lbaas08:05
*** velizarx has joined #openstack-lbaas08:08
*** ktibi has joined #openstack-lbaas08:42
openstackgerritCarlos Goncalves proposed openstack/octavia-tempest-plugin master: Fix the scenario parent job to use the live base
openstackgerritCarlos Goncalves proposed openstack/octavia-tempest-plugin master: WIP: Add octavia-v2-dsvm-py35-scenario-fedora-latest job
dollyHi all, does anyone knows who Adam Harwell is, just curios if he hangs around here ? I was looking for some input on my comment on the following patch-set,
*** ktibi has quit IRC09:19
*** numans is now known as numans_afk09:23
*** luksky has quit IRC10:40
*** luksky has joined #openstack-lbaas10:42
*** numans_afk is now known as numans10:53
cgoncalvesdolly, Adam is rm_work11:18
dollycgoncalves, Oh, coolers! thanks!11:19
*** Emine has quit IRC11:23
*** Emine has joined #openstack-lbaas11:25
*** yboaron_ has quit IRC12:32
*** yboaron_ has joined #openstack-lbaas12:32
*** yboaron_ has quit IRC13:14
*** yboaron has joined #openstack-lbaas13:15
*** velizarx has quit IRC13:20
*** yboaron_ has joined #openstack-lbaas13:49
*** yboaron has quit IRC13:52
openstackgerritMichael Johnson proposed openstack/octavia-tempest-plugin master: Fix tests to honor Octavia API versioning
*** fnaval has joined #openstack-lbaas14:09
*** luksky has quit IRC14:31
*** reedipb has joined #openstack-lbaas14:35
jitekaHello, I have a question regarding spare pool feature and the fact that it's not compatible with anti-affinity at the moment14:42
jiteka1. Are you currently discussing possible solution to have both compatible ?14:42
jiteka2. If I still decide to keep using ACTIVE_STANDBY topology with anti-affinity, and end-up in case where both MASTER and BACKUP are on the same compute,14:42
jitekaIs it possible to disable spare pool then failing over one of the amphora to ensure anti-affinity ?14:42
*** yboaron_ has quit IRC14:43
*** dolly has quit IRC14:50
*** Swami has joined #openstack-lbaas14:50
johnsomjiteka I believe that if you have spares pool enabled it simply will not be used for failover.  Basically we didn't want to get into the scheduler business and nova doesn't handle "adding" booted instances to an anti-affinity server group.14:52
jitekajohnsom: spare pool is only used for new loadbalancer ?14:55
johnsomIn active/standy topology mode, I think that is correct14:56
johnsomBut it might be disabled there too due to the same issue adding booted instances to the server group14:56
johnsomIt's been a while since I have looked at that code.14:56
jitekajohnsom: I'm gonna perform some testing on that as I want to see if I can use spare pool and still have some Ops way to remediate to VIP with degraded HA due to MASTER and BACKUP living on the same compute14:57
johnsomjiteka You could open a Nova RFE to update the server group anti-affinity to support live migration or some other way of adding a node to a existing server group. If we had a way to do that reliably without requiring us to become a scheduling engine it would be best.14:59
johnsomjiteka I think to date, we haven't had anyone that truely needed the performance of spares pool when running active/standby, so it hasn't been a focus for us.14:59
*** celebdor has quit IRC15:00
johnsomjiteka Would be good to get a story in for Octavia too, with your use case. Maybe we can get creative and go through the spares pool attempting to add them to the server group. I'm not sure if Nova will give us a useful exception that we could attempt to keep trying spare instances or not. Maybe it would just cycle through the list and force a boot anyway.15:05
jitekajohnsom: unfortunately right now live-migration is not an option as we rely on flavor with vCPU pinning and NUMA, and nova-scheduler is not aware of NUMA topology and vCPU pinning when live-migrating on a targeted host leading to overlap vCPU and degraded performance15:08
johnsomSo when you allocate a vCPU you actually mean it... ha15:09
jitekajohnsom: that's why performing regular audit on amphora fleet to solve LB not fullfilling anti-affinity would do the trick15:09
jitekajohnsom: will do some testing in devstack and in our pre-prod environment then see my options15:10
jitekajohnsom: thanks for the precision :)15:11
xgerman_jiteka: what you want sounds like soft-anti-affinity15:11
jitekaxgerman_: I saw that option but I don't really understand the difference between "soft" and "normal" anti-affinity15:11
xgerman_we “could” add a case if you specify soft that we use the spare pool for failover knowing you might end up on the same server15:11
johnsomThey are nova terms. xgerman_ is probably describing them as I type this....15:12
jitekaxgerman_: I had the feeling that soft-anti-affinity is a best effort as it will try to ensure anti-affinity without failing the build with NoValidHost exception if there is no available hypervisor that would be compliant with that filter15:12
xgerman_yeah, and we *could* enhance Octavia to use spare pool if you specify soft — since then you signal you tolerate being on the same “server"15:13
johnsomYeah, soft anti-affinity is "best effort" to not co-locate the amps on the same host, but if there is no other option in the cloud, go ahead and schedule the nova instance to the same compute host.15:14
xgerman_yep, which would make using spare pool legit15:14
xgerman_though probably not what jiteka wants ;-)15:15
jitekaxgerman_: indeed, it's not exactly what I have in mind as HA and ensuring that losing an hypervisor won't end-up with a VIP going down because both amphora were running here, sounds to me like good Ops practice15:16
rm_workcgoncalves: ah thanks :P missed dolly I guess15:17
jitekaxgerman_: ideally having some LB flavor and choosing between, expendable LB that tolerate same host and downtime VS Critical LB that need to stay reachable at any time15:17
jitekaxgerman_: would be better15:17
xgerman_well, that plays with our Stein goal of flavors ;-)15:18
rm_workyeah that basically sounds like not having any anti-affinity15:19
rm_workbecause it will basically just be random15:19
rm_workwhich is ... exactly as good as not having it at all15:19
jitekais it possible to have 2 different octavia provider ? (with different configuration)15:20
jitekaI mean 2 loadbalancing provider15:20
rm_worki believe so15:21
jitekawhich both would be octavia but with 2 separated control plane15:21
rm_workoh hmmm15:21
rm_worki'm ... not actually positive how that'd work15:21
rm_worki guess i'm thinking about flavors15:22
jitekarm_work: was thinking about that when you don't have flavor but it solve the same problem15:22
johnsomYeah, might just be worth waiting for flavors. I think you will run into DB problems trying to run two Octavia amphora based provider drivers15:24
*** spartakos has joined #openstack-lbaas15:24
rm_workI prefer AZ anti-affinity anyway :)15:31
*** pcaruana has quit IRC15:42
jitekarm_work: Stein is estimated to be released in april next year right ?15:45
rm_workyes maybe?15:46
jitekarm_work: nvm I found it here
jitekaApr 08 - Apr 12 (Stein release)15:47
jitekathanks all for the help15:47
rm_worknp ;)15:50
rm_worksuch help lol15:51
rm_work"uhhhhhhhh i dunno maybe" :P:P:P15:51
*** luksky has joined #openstack-lbaas15:55
*** ramishra has quit IRC15:57
*** Swami has quit IRC15:58
*** pcaruana has joined #openstack-lbaas16:01
*** luksky has quit IRC16:15
openstackgerritMerged openstack/octavia-tempest-plugin master: Fix the scenario parent job to use the live base
*** ccamposr has quit IRC16:51
*** spartakos has quit IRC17:11
jitekarm_work: with my question I needed to see what are my option for Stability vs Provisionning time17:12
jitekarm_work: now I know more :)17:13
*** reedipb has quit IRC17:14
*** ChanServ sets mode: -rf 17:16
*** LutzB has joined #openstack-lbaas17:16
*** ptoohill1 has joined #openstack-lbaas17:16
rm_workyeah, i think really it's fine to have no spares pool if your cloud isn't SUPER slow17:19
rm_work~1m boot time should be achievable and then whatever17:20
rm_workyeah, 1m to create a LB isn't GREAT but it's not horrible17:20
rm_workand rebuilding a backup at ~1m is fine17:20
rm_workthough, AZ-anti-affinity is :P:P17:20
rm_workactually i had intended to do host internally as well17:21
rm_workbut it's a little more complex, and probably no one would want to actually include THAT low level in octavia17:21
*** pcaruana has quit IRC17:27
*** spartakos has joined #openstack-lbaas17:35
*** links has joined #openstack-lbaas17:38
*** links has quit IRC17:38
*** irenab has quit IRC17:47
*** irenab has joined #openstack-lbaas17:50
*** pcaruana has joined #openstack-lbaas18:02
*** links has joined #openstack-lbaas18:02
*** openstackgerrit has quit IRC18:05
*** links has quit IRC18:11
*** spartakos has quit IRC18:15
*** openstackgerrit has joined #openstack-lbaas18:37
openstackgerritAndreas Jaeger proposed openstack/octavia master: Use openstack-tox-cover template
*** Swami has joined #openstack-lbaas18:37
openstackgerritGerman Eichberger proposed openstack/octavia master: [WIP] Switch amphora agent to use privsep
*** spartakos has joined #openstack-lbaas19:30
*** pcaruana has quit IRC19:35
*** Emine has quit IRC19:37
*** Emine has joined #openstack-lbaas19:38
openstackgerritMichael Johnson proposed openstack/octavia-tempest-plugin master: Disable KVM at limestone (again)
openstackgerritMichael Johnson proposed openstack/octavia-tempest-plugin master: DNM: Testing bionic nodes
openstackgerritMichael Johnson proposed openstack/octavia master: Disable KVM at limestone (again)
openstackgerritMichael Johnson proposed openstack/neutron-lbaas master: Exclude limestone from running with kvm
logan-:( did it break again johnsom?20:00
johnsomYeah, the latest ubuntu kernel regressed the issue.20:01
johnsomI'm trying bionic nodepool hosts. I'm hoping the new kernel in bionic will be "fixed" and we can just switch over to that20:01
johnsomIt's clear it's the nodepool kernel now though, so that is progress20:02
xgerman_yeah, tried a bionic devstack and that did not work on my vmware20:05
johnsomHmm, yeah, I haven't had time to try that out yet.  I just put up a job with the tempest plugin running under bionic to see what happens20:06
*** dims has quit IRC20:55
openstackgerritMerged openstack/octavia-tempest-plugin master: Fix tests to honor Octavia API versioning
*** luksky has joined #openstack-lbaas21:06
*** dims has joined #openstack-lbaas21:11
*** fnaval has quit IRC21:30
*** luksky has quit IRC22:01
* colin- waves to logan- 22:21
*** rcernin has joined #openstack-lbaas22:29
logan-o/ colin-!23:16
openstackgerritMichael Johnson proposed openstack/octavia master: Raise minimum coverage to 90%
*** Swami has quit IRC23:48
openstackgerritMichael Johnson proposed openstack/octavia master: Fix health manager performance regression

Generated by 2.15.3 by Marius Gedminas - find it at!