Earlier  
Later
New Search
Posted Nick Remark
#openstack-sdks - 2018-11-02
13:38:35 mordred dtantsur: :)
13:38:49 mordred dtantsur: fwiw - I gave up on distro python even and started using pyenv for python
13:39:18 dtantsur mordred: let us make devstack give up on distro python, and I'll be a much happier man
13:39:30 dtantsur :)
13:39:49 mordred I keep mildly floating the idea that we should just use pyenv - but I don't actually want to do the work to champion that :)
13:40:28 dtantsur I'd be fine with venv as well.. even if it just one huge venv with all services.
13:41:01 dtantsur another problem with us using the distro python is that it heavily screws 'yum update' by overwriting half of system python packages
13:43:31 mordred yes
13:43:56 mordred dtantsur: have you played with pyenv? it's really cool
13:44:03 mordred mordred@solace:~/src/git.openstack.org/openstack/openstacksdk$ pyenv versions
13:44:04 mordred system
13:44:06 mordred * 2.7.14 (set by /home/mordred/src/github.com/pyenv/pyenv/version)
13:44:08 mordred * 3.5.5 (set by /home/mordred/src/github.com/pyenv/pyenv/version)
13:44:30 dtantsur I have, yes. I don't have a pressing need for it on my laptop since fedora has plenty of python versions out of the box :)
13:44:40 mordred fair :)
13:45:20 dtantsur for centos - yes. I need to come up with some playbooks to configure it (I have a rule of never configuring my test environments by hand, only with these playbooks I linked above)
13:45:37 mordred that is an excellent rule
14:08:27 mordred Shrews: I made a few more patches related to nodepool/task-manager - https://review.openstack.org/#/c/614834/ and its ancestors
14:08:49 mordred the three leading up to it are just extracting some smaller pieces from the "remove task manager" patch
14:09:01 mordred the one I linked to is the one with a whole new thing in it
14:23:03 openstackgerrit Merged openstack-infra/shade master: Update the min version of tox to 2.0 https://review.openstack.org/615056
14:24:16 Shrews mordred: *nod*
14:31:42 openstackgerrit Merged openstack/openstacksdk master: Update min tox version to 2.0 https://review.openstack.org/615037
15:01:35 openstackgerrit Merged openstack/os-client-config master: Update min tox version to 2.0 https://review.openstack.org/615040
15:01:36 openstackgerrit Merged openstack/keystoneauth master: fix wrong spelling of "unnecessary" https://review.openstack.org/614178
15:47:43 kmalloc mordred: +1 on the rate limiting, going through the code one more time
15:48:00 kmalloc mordred: as you'd expect... because the whole KSA contract of "we wont break you... ever"
15:48:11 kmalloc this is a pretty big change in additional functionality
15:49:16 kmalloc dtantsur: i never understood why RHEL/Fedora doesn't do the user/local bit for pip where debian did. Overwriting the distro python seems like a bad idea.
15:49:30 dtantsur yeah, no idea either
15:49:43 dtantsur I just got used to add --user to any pip commands I execute
15:50:01 kmalloc it is a major reason i run ubuntu (*looks sideways at employer reference*) locally
15:50:23 kmalloc --user is good, except when it just doesn't work.
15:50:40 kmalloc or when i need to install things like 20 times because different users....
15:51:02 kmalloc though these days i just run my python envs in docker containers
15:51:20 kmalloc it might even be worth utilizing docker containers for python versions within devstack.
15:51:45 kmalloc since it really doesn't take a ton of time to build if we have a local artifact repo
15:51:49 kmalloc local-ish
15:52:10 kmalloc but pyenv seems like way less work
15:53:33 mordred kmalloc: for zuul's container builds, we're just using the python base image from docker (although we're using python:alpine)
15:54:22 kmalloc mordred: i have so many concerns about alpine... but that is a different conversation
15:54:25 mordred kmalloc: and we're using the pbrx tool I wrote - which will make you container images of things using bindep and pbr without any additional in-repo config
15:55:02 kmalloc mordred: yeah i've been looking at pbrx it's awesome
15:55:04 mordred kmalloc: well, in this case it's because of three things - bmw was already using alpine containers in production for zuul, modern bubblewrap is already in alpine - no no need for ppas, and it's smaller
15:55:11 mordred kmalloc: but yeah
15:55:45 kmalloc mordred: oooh ooh i know lets use GENTOO! /s [nothing against gentoo, just not in this case]
15:55:51 mordred hahaha
15:57:07 kmalloc i have been working to use alpine locally, but i need to rip apart my network and re-build it first because i lost access to control things via vlans....
15:57:23 kmalloc so some infrastructure bits for a local virt stack can't be stood up
15:57:36 kmalloc don't get me wrong, I like alpine's lightweight nature.
15:57:55 kmalloc I feel like it's a little too fast-and-loose with what is added to it.
15:58:15 kmalloc somewhere between alpine and debian is where I'd like to be. (Ubuntu is not the answer)
15:58:28 kmalloc debian stable*
16:46:56 kmalloc mordred: +2 on rate limit in KSA
16:47:09 kmalloc it feels like it really belongs there as much as I wish it could be higher up.
16:47:14 mordred kmalloc: woot. (pending tests, of course)
16:47:17 kmalloc and that is due to retry logic
16:47:18 kmalloc yes.
16:47:25 mordred kmalloc: yah. the retry logic is the real kicker
16:47:30 kmalloc must have tests before +A.
16:47:34 mordred kmalloc: sort of like why discovery needed to be there too
16:47:43 kmalloc functional tests are fine (doesn't need to be unit)
16:47:44 mordred since it needes to be there for keystone _anyway_ for auth to work :)
16:47:57 kmalloc it needs to be tested so we don't break
16:48:10 kmalloc unit tests of that is going to be hard(tm)
16:48:12 mordred oh - yeah. seriously
16:48:22 mordred I *think* I have an idea of how to unit test it too
16:48:55 kmalloc cool
16:49:05 kmalloc but like i said, i'm happy with that simply being functional
16:49:25 kmalloc you *might* be able to get away with unit tests utilizing freezegun
16:49:46 kmalloc and stepping forward on ticks and a very low rate (e.g. 1 per multiple seconds)
16:50:11 kmalloc concurrency could be tested via just objects that explicitly manage when they take action.
16:50:17 kmalloc but functional might be so much easier
17:00:34 Shrews mordred: comment on https://review.openstack.org/614832
17:01:53 openstackgerrit Dean Troyer proposed openstack/python-openstackclient master: Add --name-lookup-one-by-one option to server list https://review.openstack.org/568345
17:49:10 dtroyer smcginnis: thanks for all that cleanup, I wish we could merge it today :)
17:50:52 dtroyer I'm thinking to release master, rocky and queens early-mid next week…
18:00:33 openstackgerrit Merged openstack/python-openstackclient master: Address issues from volume backend commands https://review.openstack.org/610161
18:03:07 cdent dtroyer: see you in berlin?
18:10:20 dtroyer cdent: yes, I land Sunday
18:10:32 cdent sweet
18:34:37 mordred Shrews: well, clearly I'm just stupid
18:36:59 smcginnis dtroyer: No worries. I'll try to keep those cleanup's rebased as conflicts come up so they are ready to go when the time is right.
18:38:16 dtroyer smcginnis: there is an osc4 feature branch that has some work in it already. I need to merge master back in to that, but after that is done we could just merge those into the feature branch
18:38:59 smcginnis dtroyer: Yeah, that makes sense if we have a whole batch of major-version inducing changes.
18:39:30 smcginnis If you remember, let me know when a good point would be for that and I can rebase those over to the feature branch.
18:47:07 dtroyer will do. I keep threatening to do it tomorrow :)
18:48:07 openstackgerrit Merged openstack/python-openstackclient master: Make use of keystoneauth service-type filtering for versions https://review.openstack.org/585030
19:32:44 Shrews mordred: yeah, well i didn't know if that was some sort of python trickery i wasn't aware of, or... your reason :-P
19:33:06 mordred Shrews: definitely my reason
19:41:25 openstackgerrit Merged openstack/python-openstackclient master: Add --key-name and --key-unset option for server rebuild API. https://review.openstack.org/576422
22:28:33 openstackgerrit melanie witt proposed openstack/python-openstackclient master: Handle multiple ports in AddFloatingIP https://review.openstack.org/615353
22:47:45 openstackgerrit melanie witt proposed openstack/python-openstackclient master: Handle multiple ports in AddFloatingIP https://review.openstack.org/615353
23:17:35 openstackgerrit Merged openstack/osc-lib master: add lib-forward-testing-python3 test job https://review.openstack.org/594285
#openstack-sdks - 2018-11-03
14:16:30 openstackgerrit Merged openstack/openstacksdk master: Remove mocking workaround from adapter https://review.openstack.org/614831
#openstack-sdks - 2018-11-04
10:48:40 openstackgerrit Nguyen Hai Truong proposed openstack/openstacksdk master: [Trivial Fix] Correct spelling error of "bandwidth" https://review.openstack.org/615421
14:29:45 openstackgerrit Monty Taylor proposed openstack/openstacksdk master: [Trivial Fix] Correct spelling error of "bandwidth" https://review.openstack.org/615421
14:38:29 openstackgerrit Monty Taylor proposed openstack/openstacksdk master: Shift swift segment async code out of adapter https://review.openstack.org/614832
14:38:30 openstackgerrit Monty Taylor proposed openstack/openstacksdk master: Remove unused Task classes https://review.openstack.org/614833
14:38:30 openstackgerrit Monty Taylor proposed openstack/openstacksdk master: Collect request stats with statsd https://review.openstack.org/614834

Earlier   Later