| opendevreview | Jens Harbott proposed openstack/project-config master: Drop project definitions for some x/* repos https://review.opendev.org/c/openstack/project-config/+/925075 | 04:56 |
|---|---|---|
| *** tobias-urdin|pto is now known as tobias-urdin | 06:31 | |
| *** bauzas_ is now known as bauzas | 06:52 | |
| *** bauzas_ is now known as bauzas | 07:24 | |
| *** bauzas_ is now known as bauzas | 07:56 | |
| *** bauzas_ is now known as bauzas | 08:21 | |
| *** bauzas_ is now known as bauzas | 10:04 | |
| simondodsley | Hi - we are currently rbuilding our 3rd party CI on new hardware and wanted to ensure we optimize the size of the VMs we use to run the CI jobs on. Is there a preferred CPU core count and RAM size for these devstack nodes? | 14:21 |
| fungi | simondodsley: openstack runs most jobs (including devstack+tempest) on cloud virtual machines with 4-8 vcpus (depending on their speed/vintage), 8gb ram, and 80gb of disk space | 14:44 |
| simondodsley | fungi: thanks | 14:44 |
| fungi | you should aim for sufficient performance that devstack/tempest jobs complete in around 90 minutes (depending on which tests you enable) | 14:45 |
| opendevreview | Merged openstack/project-config master: Add openstack/os-test-images project under glance https://review.opendev.org/c/openstack/project-config/+/925043 | 14:50 |
| *** bauzas_ is now known as bauzas | 16:01 | |
| opendevreview | Merged openstack/project-config master: Set linaro cloud's max servers to 0 https://review.opendev.org/c/openstack/project-config/+/925029 | 16:30 |
| opendevreview | Merged openstack/project-config master: Use the osuosl mirror for deb packages in image builds https://review.opendev.org/c/openstack/project-config/+/925048 | 16:30 |
| dansmith | fungi: clarkb: any suggestions for some job definition I can look at (read: copy) that can just install some packages (qemu-img) and run a minimal python script to generate these images? | 16:38 |
| clarkb | dansmith: the tox (and nox) jobs will install anything listed in bindep then run the specified tox target. That might be a good example since you're using python to drive things | 16:39 |
| dansmith | okay I can surely do that, I just figured (ab)using tox for that might be a bit uncool, | 16:40 |
| dansmith | but I guess we use it for docs | 16:40 |
| clarkb | its helpful when running python tools as it ensures you can get all the deps installed too | 16:40 |
| dansmith | yeah | 16:40 |
| clarkb | you can also have a job that is basically just an ansible playbook with a single task that is a shell script that installs the pacakge and runs your commands | 16:44 |
| clarkb | but it is probably less effort to just use tox since writing a rule for that in a tox.ini is simple | 16:44 |
| opendevreview | Merged openstack/project-config master: Drop project definitions for some x/* repos https://review.opendev.org/c/openstack/project-config/+/925075 | 16:45 |
| sean-k-mooney | clarkb: im not sure where to bring this up. maybe the list but we have a depencoy on a project that is archievd on github with no future maintainer identifed and it does not work for me properly on 3.12... | 16:52 |
| sean-k-mooney | specifically https://github.com/al45tair/netifaces | 16:53 |
| sean-k-mooney | do you know if that has been discussed anywhere before | 16:53 |
| clarkb | I want to say there were rumblings of it a while back (maybe when it was archived?) so not in the context of python version support but just concern that it wasn't maintained any longer | 16:54 |
| sean-k-mooney | ya so this is only failing for me on debian testing at least right now | 16:54 |
| sean-k-mooney | but there have been no releases in 3 years | 16:55 |
| sean-k-mooney | and it was arcived arount that time too | 16:55 |
| clarkb | I think glean manually scans through the interfaces and doesn't use netifaces | 16:57 |
| sean-k-mooney | anyway for now ill swap to a ubuntu vm but we proably need to treat this like the eventlet matiannace and start removign it | 16:57 |
| clarkb | but its use case if far simpler (just needs to filter out interfaces that aren't "real") | 16:57 |
| sean-k-mooney | we use it for a few things | 16:57 |
| sean-k-mooney | in os-vif i think its only used for windows (need to check) but we remove hyperv supprot like a year ago | 16:58 |
| sean-k-mooney | so i think we can just remove it | 16:58 |
| sean-k-mooney | im not sure tha tnova actully uses it anymore (we deleted nova-net a long time ago) | 16:58 |
| sean-k-mooney | neutron uses it a little i belvie but not sure | 16:58 |
| sean-k-mooney | ok nova has some really minimal usage | 17:00 |
| sean-k-mooney | just to get the defautl ip | 17:00 |
| sean-k-mooney | so we can likely replace that | 17:00 |
| sean-k-mooney | oh so oslo_utils has a common implemation we can use instaed | 17:05 |
| clarkb | I was going to say can probably just copy relevant bits of netifaces into the various code bases that need it, but it is written in C which may be more pain than it is worth | 17:05 |
| sean-k-mooney | its not quite a 1:1 dropin | 17:08 |
| sean-k-mooney | but i think its likely close enouch | 17:08 |
| sean-k-mooney | nova is checkign that if you provide an ip for my_ip that that ip is on any interface on the host | 17:09 |
| sean-k-mooney | we could proably put that into oslo however instead of just doign that in nova | 17:11 |
| sean-k-mooney | oslo is useign netifaces to for some thigns | 17:11 |
| sean-k-mooney | anyway not todays problem but we likely need to add that to the list of techdebt to dealwith | 17:12 |
| clarkb | yes particularly if it isn't a simple replacement with another library or a sysfs lookup | 17:14 |
| sean-k-mooney | well we could use pyrout2 probably | 17:19 |
| sean-k-mooney | its a depency of os-vif which nova depends on | 17:19 |
| sean-k-mooney | but im not 100% sure it does this | 17:19 |
| sean-k-mooney | i think listing all the adresses however is someting it can do | 17:19 |
| sean-k-mooney | its bascially just parsing ip -a | 17:19 |
| sean-k-mooney | but we obvioulsy dont want to just shell out if we can avoid it | 17:20 |
| sean-k-mooney | im sure this is in /sys/class/net/* | 17:20 |
| sean-k-mooney | or similar but thats not gerneally a sable interface | 17:20 |
| sean-k-mooney | so we avoid using sysfs if we have other options | 17:20 |
| fungi | for reference, here was the openstack-discuss thread from the time netifaces was first archived: https://lists.openstack.org/archives/list/openstack-discuss@lists.openstack.org/thread/FFGAZ7VAR44ZD7FPPLBUOAI65JQKUD4V/#QB5T5W6TWN3SCO7AVCGS5AA2IESVHYE2 | 17:27 |
| clarkb | I suspect that sysfs is stable enough for this though. If you read the kernel docs on this they say things like dont' depend on specific error codes and that device order can change. But I suspect the actual attributes around things like what is my ip address on this interface is pretty stable | 17:28 |
| sean-k-mooney | https://stackoverflow.com/a/54629245 that seams like a nice clean approch | 17:29 |
| sean-k-mooney | just using socket's getaddrinfo | 17:29 |
| sean-k-mooney | no external deps | 17:29 |
| sean-k-mooney | addied in 3.2 and extended in 3.7 for som ipv6 usecases | 17:30 |
| sean-k-mooney | so it shoudl be fine on all supported versions | 17:30 |
| sean-k-mooney | clarkb: the main issue with sysfs is finding where its sotred and ensuring there is no divergence acrosos bonding/tunnels ectra | 17:32 |
| sean-k-mooney | its much more preferabel to use a ptyhon lib for this if possibel | 17:32 |
| sean-k-mooney | i would generally perfer to take a depency on "ifconfig" or "ip a" unless there is a portabel way to do thsi for sysfs that is docuemetned well | 17:34 |
| fungi | does the kernel's "don't break userland" mantra not extend to sysfs interfaces? | 17:35 |
| sean-k-mooney | in general no | 17:35 |
| fungi | that's too bad | 17:35 |
| sean-k-mooney | the probelem is that much of sysfs is defied by the device drivers | 17:35 |
| clarkb | its more nuanced than that | 17:35 |
| sean-k-mooney | it is | 17:36 |
| clarkb | its basically the stuff that is considered kernel internal like errors can change | 17:36 |
| clarkb | as well as order and probably content in driver specific stuff | 17:36 |
| clarkb | anyway glean has done things like https://opendev.org/opendev/glean/src/branch/master/glean/cmd.py#L1305-L1319 and https://opendev.org/opendev/glean/src/branch/master/glean/cmd.py#L1235-L1237 for years and its been fine as far as I can tell | 17:36 |
| sean-k-mooney | well os-vif does use sysfs to deal with how mellonox report virtual funcitons | 17:36 |
| sean-k-mooney | and we had to update it 2 in the last 6 years | 17:37 |
| sean-k-mooney | because they broke the abi | 17:37 |
| sean-k-mooney | btu if i can steals workign code form glean then im not agasint that :) | 17:37 |
| clarkb | the carrier attribute and 1 being live is probably something that is well understood. The more you get into driver specifics the more likely things will break | 17:37 |
| sean-k-mooney | clarkb: so i was trying to find where the ip is locally since adress is actully the mac | 17:38 |
| sean-k-mooney | i dont see glean actully geting the ip | 17:40 |
| clarkb | sean-k-mooney: no as I mentioend its just filtering interfaces it doesn't want to deal with | 17:41 |
| clarkb | so basically it needs to know if the interface is alive and if it is of a valid type | 17:41 |
| sean-k-mooney | ack ya thats not why we use netifaces | 17:41 |
| sean-k-mooney | we are suign it for 2 thigns 1 to get our default ip and to if you provide one to ensure its on the current host | 17:42 |
| sean-k-mooney | we used it for other thigns on windows but we dont supprot that in nova/neutron anymore | 17:42 |
| sean-k-mooney | anyway thanks, didntn want to neard snip ye into figuring out how to fix this | 17:43 |
| dansmith | clarkb: fungi JayF: https://review.opendev.org/c/openstack/os-test-images/+/925114/comments/daf9fd36_d416d4d6 | 17:44 |
| dansmith | I did away with the (good|bad) prefix so that the image names remain stable, and the manifest.yaml has the flags in it, which should be easy to parse so we know if something should work or not | 17:44 |
| dansmith | I want to go through and add a description field for each one sort of explaining why each is relevant too | 17:45 |
| opendevreview | Merged openstack/project-config master: Remove labels and diskimages from the linaro cloud https://review.opendev.org/c/openstack/project-config/+/925049 | 18:50 |
| *** bauzas_ is now known as bauzas | 20:16 | |
Generated by irclog2html.py 2.17.3 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!