| *** denghui has quit IRC | 00:01 | |
| *** denghui has joined #openstack-lbaas | 00:01 | |
| *** bharathm has joined #openstack-lbaas | 00:12 | |
| *** rm_you has quit IRC | 00:15 | |
| *** vivek-ebay has joined #openstack-lbaas | 00:26 | |
| *** vivek-ebay has quit IRC | 01:09 | |
| *** Brian_shang has joined #openstack-lbaas | 01:18 | |
| *** Brian_shang has quit IRC | 01:26 | |
| *** Brian_shang has joined #openstack-lbaas | 01:27 | |
| *** diogogmt has quit IRC | 01:28 | |
| *** diogogmt has joined #openstack-lbaas | 01:29 | |
| *** Brian_shang has quit IRC | 01:31 | |
| *** Brian_shang has joined #openstack-lbaas | 01:32 | |
| *** bana_k has joined #openstack-lbaas | 01:35 | |
| *** bank_ has joined #openstack-lbaas | 01:36 | |
| *** telmich_ has joined #openstack-lbaas | 01:36 | |
| *** telmich has quit IRC | 01:38 | |
| *** bank_ has quit IRC | 01:40 | |
| *** bana_k has quit IRC | 01:40 | |
| *** bank_ has joined #openstack-lbaas | 01:45 | |
| *** bharathm has quit IRC | 01:49 | |
| *** bank__ has joined #openstack-lbaas | 01:49 | |
| *** bank_ has quit IRC | 01:51 | |
| *** vivek-ebay has joined #openstack-lbaas | 02:27 | |
| *** ajmiller has joined #openstack-lbaas | 02:30 | |
| *** ajmiller has quit IRC | 02:31 | |
| *** vivek-ebay has quit IRC | 03:06 | |
| *** Brian_shang has quit IRC | 03:07 | |
| *** Brian_shang has joined #openstack-lbaas | 03:08 | |
| *** bank__ has quit IRC | 03:13 | |
| *** diogogmt has quit IRC | 03:24 | |
| *** denghui has quit IRC | 03:40 | |
| *** denghui has joined #openstack-lbaas | 03:41 | |
| *** bank_ has joined #openstack-lbaas | 03:53 | |
| *** amotoki has joined #openstack-lbaas | 04:27 | |
| *** vjay10 has joined #openstack-lbaas | 04:37 | |
| *** bana_k has joined #openstack-lbaas | 04:37 | |
| *** vjay10 has quit IRC | 04:44 | |
| *** vjay10 has joined #openstack-lbaas | 04:47 | |
| *** vjay10 has quit IRC | 04:51 | |
| *** bana_k has quit IRC | 04:51 | |
| *** bana_k has joined #openstack-lbaas | 04:51 | |
| *** numans has joined #openstack-lbaas | 04:52 | |
| *** bank_ has quit IRC | 04:53 | |
| *** bank_ has joined #openstack-lbaas | 04:53 | |
| *** bana_k has quit IRC | 04:57 | |
| *** bank_ has quit IRC | 04:58 | |
| *** bank_ has joined #openstack-lbaas | 05:26 | |
| *** bana_k has joined #openstack-lbaas | 05:28 | |
| *** Brian_shang has quit IRC | 05:39 | |
| *** Brian_shang has joined #openstack-lbaas | 05:40 | |
| *** vjay10 has joined #openstack-lbaas | 05:44 | |
| *** bank_ has quit IRC | 05:45 | |
| *** bank_ has joined #openstack-lbaas | 05:49 | |
| *** sbalukoff has quit IRC | 05:55 | |
| *** bana_k has quit IRC | 05:58 | |
| *** bana_k has joined #openstack-lbaas | 05:58 | |
| *** sbalukoff has joined #openstack-lbaas | 06:08 | |
| *** bank_ has quit IRC | 06:33 | |
| *** vjay10 has quit IRC | 06:35 | |
| *** bana_k has quit IRC | 06:38 | |
| *** kiran-r has joined #openstack-lbaas | 06:55 | |
| *** itsuugo has joined #openstack-lbaas | 06:56 | |
| *** vjay10 has joined #openstack-lbaas | 06:58 | |
| itsuugo | hi, I'm lost in LbaaS documentation, if I want to have LBaaS with L7 extensions what do I need? octavia, lbaasv2? | 07:11 |
|---|---|---|
| *** itsuugo has quit IRC | 07:16 | |
| *** itsuugo has joined #openstack-lbaas | 07:17 | |
| *** chlong has quit IRC | 07:17 | |
| *** vjay11 has joined #openstack-lbaas | 07:39 | |
| *** vjay10 has quit IRC | 07:39 | |
| *** vjay11 has quit IRC | 07:47 | |
| *** vjay11 has joined #openstack-lbaas | 07:57 | |
| *** kiran-r has quit IRC | 08:26 | |
| *** eezhova has quit IRC | 08:31 | |
| *** eezhova has joined #openstack-lbaas | 08:34 | |
| openstackgerrit | Elena Ezhova proposed openstack/neutron-lbaas: [WIP][LBaaS v2] Improve API tests performance https://review.openstack.org/226370 | 08:53 |
| *** kiran-r has joined #openstack-lbaas | 08:57 | |
| *** apuimedo has joined #openstack-lbaas | 09:14 | |
| Brian_shang | itsuugo: No need octavia, just use lbaasv2 with HaproxyNSDriver. | 09:26 |
| itsuugo | thxs | 09:27 |
| *** sc68cal has quit IRC | 09:52 | |
| itsuugo | Brian_shang: is lbaasv2 available in kilo? I can't find any doc to enable it :( | 09:54 |
| *** amotoki has quit IRC | 09:55 | |
| *** sc68cal has joined #openstack-lbaas | 09:56 | |
| *** sc68cal has quit IRC | 09:56 | |
| *** sc68cal has joined #openstack-lbaas | 09:56 | |
| Brian_shang | itsuugo: yes, you can read neutron-lbaas/devstack/plugin.sh to enable it. | 09:57 |
| itsuugo | I'm using ubuntu repos for openstack kilo, and seems that lbaas v2 is not working there | 10:20 |
| itsuugo | I have enabled service_provider=LOADBALANCERV2:Haproxy:neutron_lbaas.drivers.haproxy.plugin_driver.HaproxyOnHostPluginDriver:default | 10:20 |
| itsuugo | and tried with neutron.conf:service_plugins = lbaasv2 neutron.conf:service_plugins = lbaas without success | 10:21 |
| *** kiran-r has quit IRC | 10:22 | |
| *** sballe has joined #openstack-lbaas | 10:45 | |
| *** nmagnezi__ has joined #openstack-lbaas | 11:03 | |
| *** nmagnezi__ has quit IRC | 11:07 | |
| *** Brian_shang has quit IRC | 11:12 | |
| *** Brian_shang has joined #openstack-lbaas | 11:12 | |
| *** vjay11 has quit IRC | 11:16 | |
| *** chlong has joined #openstack-lbaas | 11:43 | |
| *** itsuugo has quit IRC | 11:47 | |
| *** itsuugo has joined #openstack-lbaas | 12:03 | |
| *** itsuugo has quit IRC | 12:51 | |
| *** numans has quit IRC | 13:25 | |
| *** vjay11 has joined #openstack-lbaas | 13:42 | |
| *** itsuugo has joined #openstack-lbaas | 13:59 | |
| *** itsuugo has quit IRC | 14:01 | |
| *** denghui has quit IRC | 14:34 | |
| *** vjay11 has quit IRC | 14:43 | |
| xgerman | rm_work: https://review.openstack.org/#/c/228074/ | 15:07 |
| *** sbalukoff has quit IRC | 15:15 | |
| rm_work | xgerman: looking | 15:17 |
| xgerman | thanks — it’s a spec Barbican/DNSaaS | 15:17 |
| *** TrevorV has joined #openstack-lbaas | 15:19 | |
| rm_work | yeah | 15:21 |
| rm_work | it looks fine other than the set of IFs on line 50-67 being super confusing to me | 15:21 |
| rm_work | i am not sure what most of those are really saying | 15:21 |
| rm_work | do you understand? | 15:21 |
| xgerman | I think they sort of make sense — since it seems DNSaS handles domains and subdomains | 15:23 |
| rm_work | so the first one: | 15:24 |
| xgerman | so of if you get a cert for *.rackspace.com it should be the same for r_work.rackspace.com | 15:24 |
| rm_work | If a subdomain is created and a wild card certificate is associated | 15:24 |
| rm_work | with the parent, the subdomain will have no wild card cert set. | 15:24 |
| rm_work | I have mydomain.com | 15:24 |
| rm_work | I set a wildcard for it | 15:24 |
| rm_work | I create yours.mydomain.com | 15:24 |
| xgerman | ok, I think they allow domains be managed for more than one tenant | 15:25 |
| johnsom | I'm trying to think about the relationship with anchor here. | 15:25 |
| rm_work | yours.mydomain.com will NOT have a wildcard configured (meaning, the wildcard for the top-level is explicitly not shared) ? | 15:25 |
| xgerman | yep, they are multi tenant. So the admin might own *.ebay.com and each tenant can have a sundomain | 15:25 |
| rm_work | and the wildcard for the tld is explicitly not shared lower | 15:26 |
| xgerman | yep | 15:26 |
| rm_work | so someone.yours.mydomain.com would not work | 15:26 |
| rm_work | or, it would, but it wouldn't have a cert | 15:26 |
| xgerman | yeah, which seems limiting | 15:26 |
| rm_work | though err | 15:26 |
| johnsom | rm_work while you are around, this is a simple one to get our coverage stuff going again: https://review.openstack.org/#/c/225548/ | 15:27 |
| rm_work | if I have a wildcard cert for mydomain.com | 15:27 |
| xgerman | johnsom you got a -1 | 15:27 |
| rm_work | it DOES cover *.mydomain.com | 15:27 |
| rm_work | which includes yours.mydomain.com | 15:27 |
| rm_work | because that's the point of having a wildcard cert | 15:27 |
| johnsom | rm_work ah, crumb | 15:27 |
| rm_work | but this is saying that it specifically wouldn't share it | 15:27 |
| rm_work | right? | 15:27 |
| xgerman | yep, that’s their multi-tenant thing | 15:28 |
| rm_work | which ... what is the point of having a wildcard cert at all if it isn't shared with subdomains | 15:28 |
| rm_work | johnsom: yeah i THINK we need to get that !=4.0 in global | 15:29 |
| rm_work | because otherwise our reqs could cause problems in a shared env | 15:29 |
| xgerman | +1 | 15:29 |
| johnsom | Yeah, probably. | 15:29 |
| rm_work | or, figure out why 4.0 breaks us | 15:29 |
| johnsom | I will work on that. The 4.0 release is clearly broken | 15:29 |
| rm_work | if it's a buggy release, then yeah we need !=4.0 in global and they need to release an update | 15:30 |
| johnsom | Ned the developer doesn't seem to have an idea of what is going on. | 15:30 |
| rm_work | >_> | 15:30 |
| johnsom | It seems to me this cert via dns spec is a bit dangerous. Usually you don't want people to wildly create signed subdomains, etc. There should be some diligence with issuing signed certs. | 15:38 |
| rm_work | well, there is a lot of info required for this to work | 15:39 |
| rm_work | A) it's not at all a synchronous process with Barbican | 15:39 |
| johnsom | I'm just looking at the section about multiple tenants | 15:39 |
| rm_work | B) in order to get a cert, you need to provide quite a bit of specific info | 15:40 |
| xgerman | yeah, I think they didn’t understand how Barbican works/or want to use their own CA | 15:41 |
| * redrobot pokes head in | 15:41 | |
| redrobot | xgerman what questions do you have? | 15:42 |
| rm_work | redrobot: we're looking at https://review.openstack.org/#/c/228074/1 | 15:42 |
| xgerman | yeah, my mugs over there brought it to my attention | 15:42 |
| johnsom | Line 60 down is making me scratch my head | 15:42 |
| xgerman | mugsy | 15:42 |
| rm_work | err, accidentally +1d instead of -1d lol | 15:42 |
| rm_work | johnsom: i think 60+ is the only part that makes SENSE to me | 15:43 |
| rm_work | since the "CA" is just a thing global to all of barbican | 15:43 |
| rm_work | so if you are using CA ID 3 | 15:43 |
| rm_work | and the domain transfers to another tenant | 15:43 |
| rm_work | CA ID 3 is the same CA | 15:44 |
| rm_work | why that matters, I don't quite know | 15:44 |
| *** Kiall has joined #openstack-lbaas | 15:44 | |
| rm_work | but it'd still have the same CA available in Barbican to create a new cert | 15:44 |
| xgerman | or renew the cert | 15:44 |
| *** mugsie has joined #openstack-lbaas | 15:44 | |
| rm_work | yes | 15:44 |
| openstackgerrit | Merged openstack/neutron-lbaas: Kill HEADS file https://review.openstack.org/227466 | 15:47 |
| *** mlavalle has joined #openstack-lbaas | 16:01 | |
| *** vjay11 has joined #openstack-lbaas | 16:33 | |
| openstackgerrit | Phillip Toohill proposed openstack/neutron-lbaas: Allow updating TLS refs https://review.openstack.org/220654 | 16:33 |
| *** alejandrito has joined #openstack-lbaas | 16:39 | |
| *** vivek-ebay has joined #openstack-lbaas | 16:43 | |
| *** Brian_shang has quit IRC | 16:47 | |
| *** Brian_shang has joined #openstack-lbaas | 16:48 | |
| openstackgerrit | Phillip Toohill proposed openstack/octavia: Updates for containers functionality https://review.openstack.org/199954 | 16:48 |
| *** vivek-ebay has quit IRC | 16:51 | |
| openstackgerrit | Merged openstack/octavia: Remove quotes from subshell call in bash script https://review.openstack.org/226716 | 16:52 |
| *** vivek-ebay has joined #openstack-lbaas | 17:01 | |
| *** TrevorV has quit IRC | 17:05 | |
| *** TrevorV has joined #openstack-lbaas | 17:06 | |
| *** minwang2 has joined #openstack-lbaas | 17:11 | |
| *** bana_k has joined #openstack-lbaas | 17:13 | |
| *** bank_ has joined #openstack-lbaas | 17:14 | |
| *** apuimedo has quit IRC | 17:21 | |
| openstackgerrit | Phillip Toohill proposed openstack/octavia: Updates for containers functionality https://review.openstack.org/199954 | 17:22 |
| *** apuimedo has joined #openstack-lbaas | 17:23 | |
| *** vjay11 has quit IRC | 17:23 | |
| *** vivek-ebay has quit IRC | 17:31 | |
| *** vivek-ebay has joined #openstack-lbaas | 17:32 | |
| *** sbalukoff has joined #openstack-lbaas | 17:34 | |
| *** pckizer has joined #openstack-lbaas | 17:37 | |
| *** Aish has joined #openstack-lbaas | 17:46 | |
| *** madhu_ak has joined #openstack-lbaas | 18:13 | |
| bana_k | how do i specify the session persistence while creating the pool ? | 18:20 |
| *** mestery_afk is now known as mestery | 18:22 | |
| pc-pothole | bana_k: Is it not allowing you to specifiy session persistence in the request body? | 18:25 |
| bana_k | nope its letting me allow | 18:39 |
| bana_k | but I am not able to figure it out, but I think I got it now | 18:39 |
| johnsom | FYI, the global requirements change to exclude coverage 4.0: https://review.openstack.org/#/c/228581/1 | 18:49 |
| pc-pothole | johnsom: Question: In the controller worker around line 50, the ComputeBuildException is caught and rethrown. I think I understand why this is done, but since this is also done in my review and others have questions regarding it im asking here in hopes you could clarify for us. | 18:53 |
| *** fnaval has quit IRC | 18:55 | |
| *** minwang2 has quit IRC | 18:56 | |
| *** barclaac|2 has joined #openstack-lbaas | 19:02 | |
| bana_k | In octavia.conf I see we have a flavor as 10, is it nova flavor? | 19:02 |
| *** bank_ has quit IRC | 19:03 | |
| pc-pothole | bana_k: the default that's commented out in the conf is empty for me. | 19:04 |
| pc-pothole | The retries have 10 | 19:04 |
| *** barclaac has quit IRC | 19:05 | |
| bana_k | oh I think I might have the old code. It shows me as 10 | 19:05 |
| pc-pothole | oh in the devstack settings | 19:05 |
| pc-pothole | I dont have devstack running atm, but you could query to see the flavor id's and see if they match. | 19:06 |
| bana_k | yes I see that in octaiva conf we have flavor as 10 n nova flavor-list does not have that | 19:07 |
| pc-pothole | it should be the nova flavor. the devstack plugin is setting this. You can see the default in the settings file. | 19:09 |
| *** barclaac|2 has quit IRC | 19:09 | |
| bana_k | ok, ll see | 19:09 |
| bana_k | thanks | 19:09 |
| pc-pothole | ill stack see what i get | 19:10 |
| *** barclaac|2 has joined #openstack-lbaas | 19:10 | |
| bana_k | that will be great :). thanks | 19:10 |
| *** itsuugo has joined #openstack-lbaas | 19:11 | |
| *** Aish has quit IRC | 19:20 | |
| *** fnaval has joined #openstack-lbaas | 19:23 | |
| blogan | bana_k: do a flavor-list --all | 19:25 |
| blogan | bana_k: its created as not public | 19:26 |
| pc-pothole | yea, bana_k I have m1.amphora as the flavor with ID of 10. | 19:36 |
| rm_work | it's created as part of the devstack spinup | 19:38 |
| rm_work | https://github.com/openstack/octavia/blob/master/devstack/plugin.sh#L146 | 19:38 |
| rm_work | https://github.com/openstack/octavia/blob/master/devstack/settings#L32 | 19:39 |
| rm_work | bana_k: ^^ | 19:39 |
| *** woodster_ has joined #openstack-lbaas | 19:42 | |
| johnsom | bana_k I added flavor 10 to the devstack plugin.sh to be able to cut the amount of memory down to 1GB of RAM and 2GB of disk | 19:48 |
| johnsom | pc-pothole Can you give me a file:line#? In controller_worker.py:50 is the init class with member_repo | 19:50 |
| pc-pothole | johnsom: i was looking at the wrong things. No longer an issue, sorry for that. | 19:50 |
| bana_k | oh ok. will check out johnsom, thanks | 19:51 |
| johnsom | Ok | 19:51 |
| bana_k | this is the wip https://review.openstack.org/#/c/228598/ | 19:51 |
| bana_k | for heat hooks | 19:51 |
| bana_k | please review | 19:52 |
| bana_k | It still needs a lot of work n input | 19:52 |
| johnsom | Cool | 19:52 |
| openstackgerrit | Phillip Toohill proposed openstack/octavia: Updates for containers functionality https://review.openstack.org/199954 | 19:52 |
| *** minwang2 has joined #openstack-lbaas | 19:52 | |
| *** Aish has joined #openstack-lbaas | 19:56 | |
| *** vivek-ebay has quit IRC | 20:05 | |
| openstackgerrit | Phillip Toohill proposed openstack/octavia: Adding new network driver for containers https://review.openstack.org/197858 | 20:05 |
| openstackgerrit | Phillip Toohill proposed openstack/octavia: Updates for containers functionality https://review.openstack.org/199954 | 20:07 |
| openstackgerrit | German Eichberger proposed openstack/octavia: Explicitly kills a process listening on the amphora port https://review.openstack.org/227912 | 20:22 |
| *** mlavalle has quit IRC | 20:27 | |
| *** barclaac|2 has quit IRC | 20:38 | |
| *** barclaac|2 has joined #openstack-lbaas | 20:41 | |
| openstackgerrit | Michael Johnson proposed openstack/octavia: Fix a typo in the .coveragerc file https://review.openstack.org/225548 | 20:41 |
| *** vivek-ebay has joined #openstack-lbaas | 20:49 | |
| *** mlavalle has joined #openstack-lbaas | 20:54 | |
| openstackgerrit | min wang proposed openstack/octavia: Add cert tracking and rotating in Housekeeping https://review.openstack.org/215359 | 20:55 |
| *** Aish has quit IRC | 21:15 | |
| *** Aish has joined #openstack-lbaas | 21:17 | |
| *** itsuugo has quit IRC | 21:29 | |
| *** TrevorV has quit IRC | 21:36 | |
| *** vivek-ebay has quit IRC | 21:51 | |
| blogan | xgerman, johnsom, rm_work, dougwig: we are probably going to have to skip some tests in the api-lb job | 22:01 |
| blogan | temporarily | 22:01 |
| xgerman | k | 22:01 |
| johnsom | Why is that? | 22:01 |
| dougwig | i would say let's upgrade that probably to definitely. or split the jobs further. | 22:01 |
| openstackgerrit | min wang proposed openstack/octavia: Add cert tracking and rotating in Housekeeping https://review.openstack.org/215359 | 22:02 |
| xgerman | well. let’s see if we can get the container stuff going | 22:02 |
| * johnsom looks puzzled... We are running faster than 3-4 other jobs, what changed? | 22:02 | |
| dougwig | the lb job is going >120 minutes. | 22:03 |
| dougwig | no clue | 22:03 |
| rm_work | err | 22:04 |
| rm_work | yes | 22:04 |
| johnsom | I saw one over the weekend, but that was before our tests even ran | 22:04 |
| rm_work | err | 22:04 |
| rm_work | hmm | 22:04 |
| rm_work | 120 is the timeout? | 22:04 |
| johnsom | 125 actually. | 22:04 |
| rm_work | the MINIMAL is still fine at <60m right? | 22:04 |
| johnsom | Which one are you looking at? | 22:04 |
| rm_work | load_balancers was the long one since the beginning | 22:04 |
| rm_work | but i thought it was going <120 | 22:04 |
| *** vivek-ebay has joined #openstack-lbaas | 22:05 | |
| johnsom | Yeah, that was the long one | 22:05 |
| xgerman | so, why can’t we run our tests as long as we want? | 22:05 |
| blogan | http://logs.openstack.org/87/226587/2/check/gate-neutron-lbaasv2-dsvm-loadbalancer/bb56340/console.html | 22:07 |
| blogan | xgerman: because if they allow all the integration tests to run as long as wanted, and then they have to retrigger when something ahead of them fails, it'll get backed up to no end | 22:07 |
| blogan | bc the gate jobs will have to restart again if there's a failure ahead of them | 22:07 |
| johnsom | Yeah, those are really slow hosts | 22:08 |
| *** alejandrito has quit IRC | 22:09 | |
| johnsom | That is really strange, they are failing before the timeout. | 22:11 |
| blogan | johnsom: new mechansim got merged on friday that causes that | 22:11 |
| johnsom | This one ran until 2:01:25 https://review.openstack.org/#/c/227363/ | 22:11 |
| xgerman | ah, so mechanism need to fix their act... | 22:12 |
| xgerman | our work is done — kick them out... | 22:12 |
| xgerman | kidding... | 22:12 |
| blogan | johnsom: so the timeout is recalculated a bit now, the timeout doesn't include the devstack build so there is definitely some time changes | 22:12 |
| johnsom | The variability in the hosts is what worries me, this one finished in 1:30:49 https://review.openstack.org/#/c/227466/ | 22:12 |
| johnsom | blogan patch link? | 22:13 |
| johnsom | I'm lost on that one too, as without the devstack setup time you patch test ran 1:21 ish | 22:14 |
| blogan | https://review.openstack.org/#/c/227476/ | 22:14 |
| *** vivek-ebay has quit IRC | 22:17 | |
| johnsom | I like that there is a chance we could get a full log set on timeout, but it is not clear how much time they took away from the test itself. | 22:18 |
| johnsom | timeout -s 9 81m bash | 22:21 |
| johnsom | So, tests now have to run inside 81 minutes | 22:22 |
| johnsom | depending on how long the host takes for devstack setup | 22:22 |
| johnsom | Ugh, ugly | 22:22 |
| blogan | yeah i know | 22:22 |
| johnsom | I would have blocked this for nested virt if it was clear the impact. | 22:23 |
| johnsom | (not that my vote counts) | 22:23 |
| blogan | well do any of us pay attention to all reviews in all of these projects? | 22:24 |
| xgerman | well, don’t they have to pass jenkins before? | 22:25 |
| xgerman | so if we time out and therefore vote no how do they get in? | 22:25 |
| rm_work | gate changes don't test against projects | 22:25 |
| xgerman | ah, ok | 22:26 |
| rm_work | which is why normally you make gate stuff experimental first, then test, then flip to normal/voting | 22:26 |
| xgerman | yeah, that’s how we do it | 22:26 |
| johnsom | So, on a good host we take 1:08:17 and on a bad host we took 1:15:03 | 22:28 |
| johnsom | I guess we should aim to take about 15 minutes out, which is roughly three tests. | 22:29 |
| xgerman | well, then the next change comes… I think we have a systemic problem | 22:30 |
| johnsom | Yep | 22:31 |
| johnsom | Easy solution is to split the load balancer tests in half | 22:32 |
| xgerman | ok, we should do that | 22:33 |
| xgerman | but then something happened and we need to split in 4ths | 22:33 |
| xgerman | dougwig we need some process change so others can’t break our tests that easy | 22:33 |
| blogan | what if we changed all the api tests to use the noop drivers, but then we also ran the minimal tests job we have for neutron using the VM drivers? | 22:35 |
| johnsom | Worst part hear is a change in devstack will cut our test time | 22:36 |
| johnsom | hear->here | 22:38 |
| johnsom | ugh. This all makes my head hurt. | 22:38 |
| blogan | well do you think the noop drivers for the api tests and only use the real drivers for a minimal test suite would be fine? | 22:38 |
| johnsom | noop to me seems like we are asking for trouble. | 22:38 |
| *** vivek-ebay has joined #openstack-lbaas | 22:38 | |
| xgerman | +1 | 22:39 |
| xgerman | some of those api tests look for real stuff | 22:39 |
| xgerman | so we would need to fix that | 22:39 |
| johnsom | But this comes back to my comment last week about re-evaluating what we are testing and where | 22:39 |
| blogan | well we are in trouble right now, adn if we have to keep skipping tests we'll be getting in more trouble | 22:39 |
| johnsom | Agreed. | 22:39 |
| blogan | the way i see it is, most of these tests are testing api validation, which is just updates to the db | 22:40 |
| johnsom | BTW, noop is not going to be a quick fix as I'm pretty sure it will require a bunch of fixes | 22:40 |
| blogan | actually | 22:40 |
| xgerman | well, they also test the driver | 22:40 |
| johnsom | Should the? | 22:40 |
| blogan | johnsom: im sure too but i think if i focus on it for a night i can get it working but that might be all sunshine and rainbows thinking | 22:40 |
| johnsom | they? | 22:40 |
| blogan | johnsom: actually this brings up another improvement we can make, we'eve talkeda bout it ebfore | 22:41 |
| blogan | updates to name and descriptions houldn't go do the drivers | 22:41 |
| blogan | but that wouldn't help these test runs out much | 22:41 |
| johnsom | So, that is the question right? Should API tests only test the API layer, in which case they should not be going to the drivers | 22:41 |
| blogan | bc its the creation that is causing the problem | 22:41 |
| xgerman | so we feel that most stuff does;t go the driver and we can easily separate that | 22:41 |
| xgerman | and api tests are mostly not exposing driver problems... | 22:42 |
| blogan | well if those tests only cared about the API then they woudln't care if something went ACTIVe | 22:42 |
| blogan | just that it got accepted | 22:42 |
| johnsom | I would then argue that API tests are not appropriate for the minimal gate, as nothing would test the plugging | 22:42 |
| blogan | and returned correctly | 22:42 |
| xgerman | johnsom +! | 22:43 |
| xgerman | +1 | 22:43 |
| blogan | i guess at this point its choosing the lesser of 5 evils | 22:43 |
| johnsom | So, we have the following test categories today: minimal (neutron gate), API, scenario, driver ??? | 22:44 |
| xgerman | well, if we need to choose between evils something is fundamentally wrong -- | 22:44 |
| johnsom | driver CI? | 22:44 |
| blogan | integrated | 22:44 |
| blogan | i guess | 22:44 |
| blogan | driver-integrated | 22:44 |
| xgerman | but I gotta run an errand so... | 22:44 |
| blogan | whatever i knwo what you mean :) | 22:44 |
| blogan | xgerman: way to leave us in a dire situation! | 22:45 |
| xgerman | sorry | 22:45 |
| blogan | xgerman: im just kidding with ya | 22:45 |
| blogan | xgerman: go do your more important stuff | 22:45 |
| johnsom | We just assign the work to him, that's all | 22:45 |
| blogan | easy! | 22:45 |
| johnsom | I would propose: | 22:46 |
| johnsom | minimal - does an end-to-end happy path scenario test | 22:46 |
| johnsom | API - checks for API completeness and functionality. I would argue this could be a dummy driver. | 22:47 |
| johnsom | Scenario - Extensive end-to-end tests. How do we pull this off without nested virt or containers? If those would even get us down in time. Maybe many gates. | 22:47 |
| blogan | well, we could do the first 2 right? fix the logging noop | 22:48 |
| blogan | then in the last one we just enable what we can for now and improve as time goes on, especially if we get containers or nested virt in | 22:49 |
| johnsom | Does that give us enough coverage? | 22:49 |
| blogan | though clarkb said they did a very unscientific test with nested virt and they didnt see any improvements, but i didnt argue with him bc i dont see how it owuldn't boot faster | 22:49 |
| johnsom | We have minimal now, I assume it still functions | 22:49 |
| blogan | johnsom: i think it gives us better coverage than right now where we're skipping a lot of tests | 22:49 |
| johnsom | I just worry about the dummy driver if we don't have good scenario tests. | 22:51 |
| blogan | well it'll at least be testing api, now obviously it won't test if a particular configuration breaks a VM or container build or haproxy update | 22:52 |
| johnsom | So, flip the coin for a minute. What is the harm in creating more gate jobs? Neutron has a ton | 22:52 |
| blogan | but if haproxy fails now, the tests still won't catch that | 22:52 |
| johnsom | Yeah, I noticed that | 22:53 |
| blogan | the tests now catch if neutron or nova fail really, the scenario tests will test haproxy updates happen correctly | 22:53 |
| blogan | more gate jobs? it feels dirty too | 22:53 |
| blogan | the way we're splitting tests seems a bit hacky, and doesn't seem like something anyone else is doing | 22:54 |
| blogan | we're splitting to get true parallelism basically | 22:55 |
| johnsom | BTW, if you have vmware workstation available, you can test the difference with and without nested virtualization enabled really easily. It's a check box for the VM. | 22:55 |
| blogan | but if we actually tried to do parallel execution of these tests at once i'm pretty sure 1) the tests are written for that and 2) memory issues | 22:55 |
| blogan | i dont have that available | 22:55 |
| *** mlavalle has quit IRC | 22:56 | |
| johnsom | Yeah, the only parallel option we have is multiple gates. We can't push the current instances much farther | 22:57 |
| johnsom | I'm just saying, neutron has 19 test gates | 22:57 |
| blogan | i can work on the noop drivers tongiht just to see how much work it'd be to get those working | 22:57 |
| blogan | or we can go for the immediate fix right now and skip tests | 22:58 |
| blogan | and then try to re-structure the jobs | 22:58 |
| johnsom | It's up to you since you're doing the work. I don't think we have anything so critical it has to be merged in the next couple of days. | 22:59 |
| blogan | to me, i think doing the minimal, API with noop drivers, and then real tests that we slowly get more coverage on is the best path forward | 22:59 |
| johnsom | by real tests, meaning setup more gates? | 22:59 |
| blogan | if thats what we have to do to get decent coverage with VM drivers | 23:00 |
| blogan | but if we do that just to run all the tests i dont see the point in having an API job run that runs those same tests with just noop drivers | 23:01 |
| blogan | to me, scenario tests that just test CUD oeprations (READ will happen organically in thsoe tests), will be fine | 23:02 |
| blogan | so 1 create update and delete method each for each resource (lb, listener, etc) | 23:02 |
| blogan | sorry test not method | 23:02 |
| blogan | as of now creating an lb with session persistence or a certain lb algorithm isn't going to fail the test even if sess pers or the algorithm doesn't get updated in haproxy | 23:03 |
| johnsom | Yeah, I am warming to this. With scenario tests, I would hope we can pack a lot more in without so many create/delete lbs | 23:03 |
| blogan | the current scenario tests that do 4 tests run in under an hour | 23:04 |
| blogan | sorry 3 tests | 23:05 |
| johnsom | Ha | 23:05 |
| blogan | but they are doing much more than the api tests | 23:05 |
| blogan | actually spinning up VMs to be web servers | 23:05 |
| blogan | to be load balanced | 23:06 |
| *** fnaval has quit IRC | 23:06 | |
| johnsom | Ok. Give it a go. I can probably help switching the gate around for driver/no-driver if you would like after the noop is working | 23:06 |
| blogan | sure, ill get it working and let ya know | 23:06 |
| blogan | if we go with the noop driver on these tests we can tell thsoe jobs not to even spin up nova or neutron correct? | 23:07 |
| blogan | well we'll need neutron | 23:07 |
| blogan | but not nova | 23:08 |
| blogan | i can test that out too | 23:08 |
| blogan | lol | 23:08 |
| blogan | alright im headed hom | 23:08 |
| blogan | i will work on this tonight | 23:08 |
| johnsom | Yeah, API doesn't need nova/neutron | 23:08 |
| johnsom | Just minimal | 23:08 |
| blogan | well we'll need neutron to load neutron-lbaas | 23:08 |
| blogan | octavia won't need it | 23:09 |
| johnsom | Yeah, sorry, just don't need to plug/unplug, etc. | 23:09 |
| johnsom | That is what I meant | 23:09 |
| johnsom | dougwig must just be hiding... | 23:11 |
| johnsom | He dropped out of the conversation early | 23:11 |
| blogan | he just likes to get the conversation started and then watch his puppets dance | 23:11 |
| johnsom | Pulls the pin and runs eh? | 23:11 |
| blogan | so eh doesn't have to do anything truly technical | 23:11 |
| blogan | lol thats a better analogy | 23:12 |
| blogan | alright bbl | 23:13 |
| johnsom | Good luck! | 23:13 |
| *** fnaval has joined #openstack-lbaas | 23:15 | |
| *** vivek-ebay has quit IRC | 23:26 | |
| *** vivek-eb_ has joined #openstack-lbaas | 23:26 | |
| *** bharathm has joined #openstack-lbaas | 23:26 | |
| *** Aish has quit IRC | 23:39 | |
| openstackgerrit | Michael Johnson proposed openstack/neutron-lbaas: Update the devstack readme.md for Octavia as ref https://review.openstack.org/228668 | 23:56 |
| xgerman | ^^ johnsom can you tag that to geo into RC2 | 23:58 |
| xgerman | dougwig wonder if you could help | 23:58 |
| johnsom | Yeah, I wondered about that | 23:58 |
| xgerman | :-) | 23:58 |
Generated by irclog2html.py 2.14.0 by Marius Gedminas - find it at mg.pov.lt!