| *** kmalloc has quit IRC | 01:10 | |
| *** yolanda__ has joined #zuul | 01:40 | |
| *** yolanda_ has quit IRC | 01:44 | |
| *** gregdek_ has joined #zuul | 02:09 | |
| *** gundalow1 has joined #zuul | 02:11 | |
| *** mordred has quit IRC | 02:12 | |
| *** gregdek has quit IRC | 02:12 | |
| *** gundalow has quit IRC | 02:12 | |
| *** smyers has quit IRC | 02:12 | |
| *** ianw_pto has quit IRC | 02:12 | |
| *** gregdek_ is now known as gregdek | 02:13 | |
| *** ianw has joined #zuul | 02:13 | |
| *** mordred has joined #zuul | 02:13 | |
| *** smyers has joined #zuul | 02:13 | |
| *** dkranz has quit IRC | 02:58 | |
| openstackgerrit | Fatih Degirmenci proposed openstack-infra/zuul master: Fix zuul home directory in zuul from scratch document https://review.openstack.org/564386 | 05:37 |
|---|---|---|
| openstackgerrit | Fatih Degirmenci proposed openstack-infra/zuul master: Remove terminal prompt from commands listed in the docs https://review.openstack.org/564387 | 05:41 |
| *** yolanda__ is now known as yolanda | 05:45 | |
| openstackgerrit | Felix Schmidt proposed openstack-infra/zuul master: Add start and end timestamp to task result in zuul_json callback https://review.openstack.org/563888 | 05:52 |
| tobiash | corvus: I like the zuul logo | 06:23 |
| tobiash | will there be t-shirts available in vancouver? | 06:23 |
| *** hashar has joined #zuul | 06:31 | |
| openstackgerrit | Felix Schmidt proposed openstack-infra/zuul master: Add start and end timestamp to task and play result in zuul_json callback https://review.openstack.org/563888 | 07:21 |
| *** jpena|off is now known as jpena | 07:35 | |
| *** threestrands_ has quit IRC | 07:52 | |
| *** ssbarnea_ has joined #zuul | 07:56 | |
| *** sshnaidm|off is now known as sshnaidm | 08:19 | |
| *** dtruong_ has joined #zuul | 08:27 | |
| *** dtruong has quit IRC | 08:28 | |
| openstackgerrit | Fatih Degirmenci proposed openstack-infra/zuul master: Use code-block for admin documentation and remove terminal prompts https://review.openstack.org/564387 | 08:47 |
| *** electrofelix has quit IRC | 08:47 | |
| openstackgerrit | Fatih Degirmenci proposed openstack-infra/zuul master: Use code-block for admin documentation and remove terminal prompts https://review.openstack.org/564387 | 08:49 |
| *** electrofelix has joined #zuul | 09:33 | |
| openstackgerrit | Fatih Degirmenci proposed openstack-infra/zuul master: Use code-block for admin documentation and terminal prompts https://review.openstack.org/564387 | 09:45 |
| openstackgerrit | Fatih Degirmenci proposed openstack-infra/zuul master: Use code-block for admin documentation and add terminal prompts https://review.openstack.org/564387 | 09:49 |
| *** gundalow1 is now known as gundalow | 10:13 | |
| *** gundalow has quit IRC | 10:13 | |
| *** gundalow has joined #zuul | 10:13 | |
| *** AJaeger has quit IRC | 10:40 | |
| *** yolanda_ has joined #zuul | 10:53 | |
| *** yolanda__ has joined #zuul | 10:54 | |
| *** yolanda has quit IRC | 10:56 | |
| *** yolanda_ has quit IRC | 10:58 | |
| *** AJaeger has joined #zuul | 10:59 | |
| *** jpena is now known as jpena|lunch | 11:00 | |
| fdegir | leifmadsen: regarding zuul from scracth document | 11:01 |
| *** threestrands_ has joined #zuul | 11:01 | |
| fdegir | leifmadsen: I have the draft for centos7 ready and I'm thinking if the document could be structured in a better way | 11:01 |
| fdegir | leifmadsen: there are quite many differences between how it is installed on centos7 so I can perhaps send a change to show the structure I have in my mind to see what you and others think | 11:02 |
| mordred | fdegir: I feel like Shrews was doing some poking around in that area too but I might be wrong - mostly mentioning to pong him for when he's awake | 11:06 |
| fdegir | mordred: thanks for the note - I will wait for Shrews to respond then | 11:07 |
| fdegir | mordred: I plan to send steps for ubuntu as well since I'll bring two zuul instances - one on centos and the other on ubuntu | 11:07 |
| mordred | fdegir: ++ that sounds great | 11:08 |
| leifmadsen | fdegir: yea I started on CentOS7 then bailed because it required a bunch of external dependencies that weren't included | 11:11 |
| leifmadsen | have at it :) | 11:11 |
| fdegir | leifmadsen: yes - tons of them | 11:14 |
| fdegir | leifmadsen: and additional repos like ius-release and bigtop | 11:14 |
| fdegir | leifmadsen: apart from that, the service files in zuul repo doesn't work with centos since zuul and stuff get installed in different locations - /usr/bin as oppose to /usr/local/bin | 11:16 |
| Shrews | fdegir: mordred: I'm done with any ZfS doc changes, and had no intention of doing any OS specific install docs (hoping someone else would contribute that), so have at it! :) | 11:45 |
| tristanC | fdegir: note that software-factory provides a zuul package for centos that has correct service files: https://softwarefactory-project.io/kojifiles/repos/sf-3.0-el7-release/Mash/rh-python35-zuul-3.0.1-1.el7.noarch.rpm | 11:48 |
| tristanC | fdegir: why do you need ius-release and bigtop for? | 11:50 |
| *** andrea06590 has joined #zuul | 11:59 | |
| andrea06590 | Hello everybody I am migrating Zuul v2 to v3 and I'd like to know what's the difference beetween launchers (v2) and executors (v3) as they both collect state, run jobs, so is there something additional executors do that launcher don't ? | 12:04 |
| fdegir | tristanC: bigtop is for zookeeper | 12:05 |
| fdegir | tristanC: and ius was for python35 if I'm not mistaken - need to check my notes | 12:06 |
| fdegir | tristanC: yes, it is python35 | 12:07 |
| andrea06590 | And my last question : is that a good practice to run merger and executor together on the same machine (as on our previous stack we run on the Zuul server : the scheduler, launcher, trigger with 2 additionnal mergers servers | 12:08 |
| andrea06590 | Thanks in advance, I am still a bit confused with the new naming (espacially executors) | 12:10 |
| *** sshnaidm is now known as notsshnaidm | 12:10 | |
| *** notsshnaidm is now known as nosshnaidm | 12:10 | |
| *** nosshnaidm is now known as sshnaidm | 12:10 | |
| tobiash | andrea06590: launchers are not existing anymore in v3 | 12:20 |
| tobiash | they were basically renamed because there is also a nodepool-launcher and that's bad for talking if you have a launcher in two related apps | 12:22 |
| fdegir | tristanC: it seems I don't need none of the stuff if I switch to sf-release-3.0.rpm which I will do once I verify stuff and reflect updates to my draft doc | 12:22 |
| fdegir | tristanC: thanks for the pointer | 12:22 |
| tobiash | andrea06590: regarding the mergers, you can do that but don't have to as each executor also acts as a merger | 12:23 |
| *** jpena|lunch is now known as jpena | 12:24 | |
| leifmadsen | fdegir: yea, I just didn't know if it was really worth even adding CentOS 7 then. It seems like a huge amount of work and a big mess to maintain. | 12:30 |
| leifmadsen | that's why I just decided to use something more easily supported/working | 12:30 |
| andrea06590 | yes i saw that but it actually act like the executor, doing almost the same job. So it is a good thing to replace my Mergers only by Mergers+Executor ? If so who does the scheduler contact my nodepool (the scheduler | 12:31 |
| andrea06590 | Thanks for your answers, my last question is more here to get who is the entry point for contacting my node pool. It will always be the scheduler ? | 12:33 |
| andrea06590 | (hope guys my english is correct enough ^^) | 12:33 |
| *** rlandy has joined #zuul | 12:35 | |
| tobiash | andrea06590: I'm not sure I correctly understood but in v3 you need one scheduler and at least one executor, you can have additional mergers if you like to enable more parallel merge operations (which is probably a good idea for larger systems) | 12:38 |
| tobiash | regarding the link from zuul to nodepool: the communication is handled via zookeeper as middleware so nodepool talks to zookeeper and the zuul-scheduler too | 12:39 |
| andrea06590 | Got it, that's what we did using v2, so executor is a kind of "replacement" for launcher | 12:41 |
| andrea06590 | Ok so scheduler is the only one to have zookeeper in the conf file, and the only one to communicate with the node pool | 12:42 |
| andrea06590 | that's what I was thinking because it would save us time to not replace our mergers behavior (maybe consolidating them with executors), but i'll have to replace the main stack (Zuul server, hosting gearman server) | 12:44 |
| andrea06590 | Thanks again for your help tobiash, I hope I better understand now :) | 12:44 |
| AJaeger | corvus, tobiash : the deprecation period for https://review.openstack.org/#/c/540890/ is over - if any of you wants to review and +2A during the next days, it would be nice. | 12:57 |
| *** threestrands_ has quit IRC | 12:58 | |
| tobiash | AJaeger: lgtm | 12:58 |
| AJaeger | thanks, tobiash ! | 12:59 |
| openstackgerrit | Merged openstack-infra/zuul-jobs master: Remove support for extensions as lists https://review.openstack.org/540890 | 13:20 |
| *** pwhalen has quit IRC | 13:32 | |
| *** JosefWells has joined #zuul | 13:34 | |
| *** pwhalen has joined #zuul | 13:35 | |
| *** pwhalen has joined #zuul | 13:35 | |
| corvus | andrea06590: it looks like you got all your questions answered, but also if it helps, we have this diagram which shows the service connections: https://zuul-ci.org/docs/zuul/admin/components.html | 13:54 |
| andrea06590 | I asked my question because of this diagram with too many connections, that make difficult for me to understand the exact workflow. Btw can you please confirm that Executor can have direct communication with nodes ? ("The scheduler must be able to connect to the ZooKeeper cluster used by Nodepool in order to request nodes. It does not need to connect directly to the nodes themselves, however – that function is handled by the Ex | 14:03 |
| *** swest has quit IRC | 14:04 | |
| corvus | andrea06590: that's good to know, we'll have to come up with a better diagram :) | 14:12 |
| corvus | andrea06590: yes, the executors need to be able to connect to the nodes (via ssh or winrm) that nodepool provides | 14:13 |
| corvus | (maybe we need a couple of diagrams) | 14:13 |
| *** ssbarnea_ has quit IRC | 14:21 | |
| pabelanger | in https://review.openstack.org/561177/ we added a new dependency on libre2-dev, was their something specific the python re library didn't do? | 14:23 |
| pabelanger | tobiash: question for you^ | 14:24 |
| Shrews | Anyone care to +3 this simple nodepool change? https://review.openstack.org/564045 | 14:31 |
| pabelanger | +3 | 14:34 |
| *** ssbarnea_ has joined #zuul | 14:36 | |
| Shrews | pabelanger: thx. oh, can you get this test fix too? https://review.openstack.org/564314 | 14:36 |
| pabelanger | sure | 14:37 |
| Shrews | sweet | 14:37 |
| *** andrea06590 has quit IRC | 14:38 | |
| openstackgerrit | Merged openstack-infra/nodepool master: Fix awkward merging of code https://review.openstack.org/564045 | 14:43 |
| tobiash | pabelanger: re2 doesn't support backrefs that could be exploited to dos zuul | 14:44 |
| tobiash | It was decided to migrate zuul to that so any new regex functionality should use that | 14:44 |
| openstackgerrit | Monty Taylor proposed openstack-infra/zuul-website master: WIP Use gatsby for static site generation https://review.openstack.org/564524 | 14:46 |
| openstackgerrit | Monty Taylor proposed openstack-infra/zuul-website master: Migrate the content https://review.openstack.org/564525 | 14:46 |
| mordred | corvus: ^^ related to email sent to mailing list | 14:47 |
| mordred | those don't work yet | 14:47 |
| openstackgerrit | Merged openstack-infra/nodepool master: Fix race in test_hold_expiration_no_default https://review.openstack.org/564314 | 14:47 |
| mordred | or, rather, they do work - but I need to update the zuul jobs | 14:47 |
| -openstackstatus- NOTICE: We've successfully troubleshooted the issue that prevented paste.openstack.org from loading and it's now back online, thank you for your patience. | 15:07 | |
| *** dkranz has joined #zuul | 15:15 | |
| Shrews | mordred: curious, what is the thing in https://review.openstack.org/536862 that requires tox 3.0? is it the "setenv = {[testenv]setenv}" part? | 15:30 |
| clarkb | Shrews: yes | 15:30 |
| clarkb | Shrews: I left a comment about it on the change | 15:30 |
| Shrews | clarkb: yeah i saw, but it wasn't clear. i can't find this documented in tox docs | 15:30 |
| clarkb | Shrews: I think it is a bug | 15:30 |
| clarkb | rather than a lack of feature | 15:31 |
| Shrews | i mean, there's this (http://tox.readthedocs.io/en/latest/config.html#substitution-for-values-from-other-sections), but that's not new | 15:31 |
| clarkb | Shrews: I seem to recall looking into this briefly when the bug was introduced to tox, let me see if I was good and filed a bug | 15:32 |
| clarkb | also tox is on github now? amazing | 15:33 |
| clarkb | https://github.com/tox-dev/tox/issues/595 someone else filed it | 15:33 |
| clarkb | wow I even pushed up code to attemtp to fix it, I don't remember that code | 15:34 |
| clarkb | Shrews: so ya it was just a bug in their ini parsing and variable substitution that I fixed and didn't remember fixing | 15:38 |
| clarkb | Shrews: and the lesson here is I should update my local tool installs when I fix the tool :) | 15:40 |
| *** AJaeger has quit IRC | 15:46 | |
| mordred | Shrews: fwiw, I was just believing clarkb that 3.0 was needed | 16:03 |
| clarkb | ya my bad on that | 16:04 |
| clarkb | I completely spaced on ever fixing this in the first place | 16:05 |
| pabelanger | tobiash: ack, thanks | 16:22 |
| *** jpena is now known as jpena|off | 16:45 | |
| mordred | clarkb, Shrews sorry - on a plane so my brain isn't fuly oxygenated - does that mean we can drop the version bump in that tox.ini change? | 16:51 |
| clarkb | mordred: I think it should be at least tox 2.8.1 | 16:51 |
| mordred | gotcha. how about I update the change just so it's accurate and doesn't force people to upgrade if they don't need to | 16:52 |
| openstackgerrit | Monty Taylor proposed openstack-infra/nodepool master: Switch to stestr https://review.openstack.org/536862 | 16:52 |
| openstackgerrit | James E. Blair proposed openstack-infra/zuul master: Cache configuration objects in addition to YAML dicts https://review.openstack.org/564061 | 16:55 |
| openstackgerrit | James E. Blair proposed openstack-infra/zuul master: Change TestMaxTimeout to not run ansible https://review.openstack.org/564562 | 16:56 |
| openstackgerrit | James E. Blair proposed openstack-infra/zuul master: Store source context on all config objects https://review.openstack.org/564563 | 16:56 |
| openstackgerrit | James E. Blair proposed openstack-infra/zuul master: Stop deep-copying job variables https://review.openstack.org/564564 | 16:56 |
| corvus | those are just 564061 again with some things that could be separate split out to make it smaller. plus one extra change to fixup some todos | 16:57 |
| corvus | yesterday 564061 passed tests; hopefully it will still do so as 3 separate changes :) | 16:57 |
| *** ssbarnea_ has quit IRC | 17:14 | |
| *** AJaeger has joined #zuul | 17:21 | |
| corvus | clarkb, mordred, tobiash: the stack up through 564061 is green and ready for review | 17:23 |
| openstackgerrit | Merged openstack-infra/zuul-jobs master: Add installation/deprecation documentation https://review.openstack.org/560463 | 17:43 |
| openstackgerrit | James E. Blair proposed openstack-infra/zuul master: Allow extra time for some ansible tests https://review.openstack.org/564572 | 17:49 |
| tobiash | corvus: what's the default wait timeout? | 17:51 |
| tobiash | regarding ^ | 17:51 |
| clarkb | mordred: if you are around I did respond to your cms thread with one thing that deserves some thought probably (there is actually a server side service you need to make it work maybe) | 17:52 |
| corvus | tobiash: 30s i think | 17:53 |
| corvus | tobiash: yeah, set in BaseTestCase in tests/base.py. used in waitUntilSettled in the same file | 17:53 |
| tobiash | ya, found it | 17:54 |
| * tobiash apologizes for being too lazy to search | 17:55 | |
| mordred | clarkb: thanks! replied (and I agree, that is definitely the area that requires some thought) | 18:09 |
| SpamapS | hrm, does anybody else know if the version of Ansible zuul is using these days supports 'var is success' as a when: condition? | 18:12 |
| SpamapS | "msg": "The conditional check 'yum_bubblewrap is success' failed. The error was: template error while templating string: no test named 'success'. String: {% if yum_bubblewrap is success %} True {% else %} False {% endif %}" | 18:12 |
| tobiash | SpamapS: oh didn't know that this exists | 18:13 |
| tobiash | SpamapS: I only find that in the 2.5 docs https://docs.ansible.com/ansible/2.5/user_guide/playbooks_tests.html#task-results | 18:17 |
| SpamapS | tobiash: yeah it mentions 2.1, so was hoping it would work :-P | 18:17 |
| SpamapS | http://docs.ansible.com/ansible/2.4/playbooks_tests.html shows |success | 18:19 |
| SpamapS | ah so does 2.3 | 18:19 |
| SpamapS | so that's what I'll use | 18:19 |
| *** ssbarnea_ has joined #zuul | 18:35 | |
| *** ssbarnea_ has quit IRC | 18:36 | |
| fdegir | tristanC: i installed zuul-* and nodepool-* from sf but service files don't seem to be included | 18:36 |
| fdegir | tristanC: from here https://softwarefactory-project.io/kojifiles/repos/sf-3.0-el7-release/Mash/ | 18:36 |
| *** ssbarnea_ has joined #zuul | 18:37 | |
| clarkb | fdegir: there should be working unit files in zuul and nodepool repos you can pull in (and the zuul docs explain how to set them up) | 18:43 |
| *** harlowja has joined #zuul | 18:46 | |
| fdegir | clarkb: yes but the installation locations are different between fedora and centos | 18:51 |
| *** xinliang has quit IRC | 18:51 | |
| fdegir | clarkb: the ones in the repos are for fedora | 18:51 |
| pabelanger | you can override them with a systemd override file | 18:51 |
| corvus | shouldn't they work for any distro? | 18:52 |
| fdegir | on fedora, nodepool/zuul go in to /usr/local/bin and on centos they go to /usr/bin | 18:52 |
| fdegir | and the unit files have /usr/local/bin | 18:53 |
| pabelanger | yah, centos you'd need to override path, something like: http://git.openstack.org/cgit/openstack/ansible-role-nodepool/tree/templates/etc/systemd/system/nodepool-builder.service.d/redhat.conf?id=5c8d047bd45208664faf6b55dd59c306054c45b5 | 18:53 |
| pabelanger | but fedora is now right for pip3 | 18:53 |
| corvus | so pip on centos installs into /usr/bin? | 18:54 |
| fdegir | btw, i am working on the installation instructions for centos, that’s why trying to find way to add as few repos as possible | 18:54 |
| fdegir | corvus: yes | 18:54 |
| fdegir | pabelanger: yes but not for centos | 18:54 |
| pabelanger | fdegir: right | 18:54 |
| fdegir | also python35 is missing on centos so i needed ius | 18:55 |
| fdegir | and for zookeeper, bigtop | 18:55 |
| pabelanger | I've dropped testing on centos in favor of fedora, to keep external dependencies down | 18:55 |
| fdegir | thats why i tried packages from sf | 18:55 |
| pabelanger | I also thing location for software collection is different then /usr/bin, i want to say /opt/python3? | 18:56 |
| clarkb | to make it even more confusing pytho2n installs on fedora go in /usr/bin too | 18:56 |
| fdegir | if i use sf, things go in to /opt/rh | 18:57 |
| pabelanger | fdegir: yah, that is it | 18:57 |
| fdegir | things work fine on centos if i bring stuff from all over the place | 18:57 |
| fdegir | so nothing is wrong | 18:57 |
| fdegir | but lots of things are not available in base repos | 18:57 |
| Shrews | i'm not sure we want to document using software factory packages in the ZfS docs | 18:57 |
| corvus | fdegir: if you want to catalogue all the things needed to get it working on centos, that would be great, and maybe we can start working on simplifying it. we can add it to the zuul-from-scratch docs, but in that case, we won't use the sf.io repos (as that wouldn't be zuul from scratch :) | 18:58 |
| corvus | Shrews: jinx | 18:58 |
| fdegir | corvus: that was what i started documenting actually | 18:58 |
| fdegir | i have the draft available already and was thinking of sending it as part of zuul from scratch | 18:58 |
| fdegir | and then tristanC mentioned sf so that's why i tried it | 18:58 |
| pabelanger | the main issue, is zookeeper, their isn't an RPM package | 18:59 |
| fdegir | pabelanger: there is one | 18:59 |
| corvus | we can also look into adjusting zfs to install things into a virtualenv, but i'd like it if we had a solution to make it easy for folks to run commands (like the zuul cli, etc) out of that venv. | 18:59 |
| fdegir | if you add apache bigtop repo | 18:59 |
| clarkb | corvus: with the direction pip10 is going that seems like a good idea in general | 18:59 |
| pabelanger | fdegir: right, should have said epel or software collections | 18:59 |
| corvus | or as clarkb keeps pointing out, it's easy to install and run the zk war, if adding a repo for it is too awkward :) | 19:00 |
| clarkb | corvus: :) I do it on my distro without a zk package :) | 19:00 |
| fdegir | https://stackoverflow.com/a/45556014 | 19:00 |
| Shrews | yeah, i run straight from war too | 19:00 |
| fdegir | ok, then i will make the docs formatted and send them for review | 19:01 |
| corvus | we've already moved the zk stuff into the distro-specific section. so if under fedora and ubuntu it's "install the distro package" and under centos it's "do a bunch of other stuff because it's centos" that's okay :) | 19:01 |
| fdegir | I can add unit files as well and perhaps move the existing ones under fedora27 and update the links | 19:01 |
| fdegir | or something like that | 19:01 |
| corvus | fdegir: cool, it sounds like we might spend a bit of time in review iterating on this, so be prepared for that, but i think that's the best way to proceed. get the current state out there and then try to make it better. | 19:01 |
| fdegir | that's fine | 19:02 |
| clarkb | corvus: upside to venv as well is we can use the same paths on all distros regardless of pip base behavior | 19:02 |
| corvus | fdegir: i'd love it if we could find a way to make just single unit files work, but if we have to have multiples, that's an option. | 19:02 |
| clarkb | corvus: venvs ^ would be one way | 19:02 |
| corvus | fdegir, clarkb: right. we just need to talk about venvs and CLI paths | 19:02 |
| fdegir | so the first version might be problematic since it focuses on getting it work on centos by using packages, etc. | 19:02 |
| fdegir | i can read what pabelanger pasted above to override stuff | 19:03 |
| pabelanger | systemd files and virtualenvs will need you to setup a PATH variable | 19:03 |
| fdegir | if not, i just share them as part of the change and we see | 19:03 |
| tobiash | corvus: I think I'm through your stack | 19:03 |
| corvus | fdegir: yeah, i think that's okay. it sounds like this is pretty complicated and it's stumped several folks so far. so just putting what you have up for review and then iterating on it as a group could be productive. :) | 19:03 |
| fdegir | will do | 19:03 |
| fdegir | thanks for the direction | 19:04 |
| tobiash | I added a few questions on 564061 | 19:04 |
| pabelanger | eg: http://git.openstack.org/cgit/openstack/windmill/tree/playbooks/templates/nodepool-builder/etc/systemd/system/nodepool-builder.service.d/override.conf.j2 | 19:04 |
| pabelanger | but does work as expect | 19:04 |
| clarkb | pabelanger: that isn't a universal rule, you need it there beacuse nodepool-builder forks dib which is installed in the venv too | 19:05 |
| corvus | tobiash: cool, thanks. we should go ahead and do the two minor improvements i think. | 19:05 |
| clarkb | pabelanger: executors forking ansible will need ti as well but others shouldn't | 19:05 |
| pabelanger | corvus: right, just nodepool-builder and zuul-executor | 19:05 |
| pabelanger | clarkb: ^ | 19:05 |
| pabelanger | corvus: sorry | 19:05 |
| Shrews | pabelanger: the sample service files for zuul and nodepool have fully hardcoded paths. We'd need to remove the path bit, yeah? | 19:06 |
| clarkb | Shrews: systemd requires rooted paths | 19:07 |
| pabelanger | Shrews: no, they should be fine. Infact systemd will want full paths to binaries | 19:07 |
| *** xinliang has joined #zuul | 19:07 | |
| *** xinliang has quit IRC | 19:07 | |
| *** xinliang has joined #zuul | 19:07 | |
| Shrews | man that's weird, but ok :) | 19:07 |
| clarkb | it is a security driven decision iirc. Basically if you run a service you know exactly what path is used always | 19:15 |
| openstackgerrit | James E. Blair proposed openstack-infra/zuul master: Cache configuration objects in addition to YAML dicts https://review.openstack.org/564061 | 20:14 |
| openstackgerrit | James E. Blair proposed openstack-infra/zuul master: Stop deep-copying job variables https://review.openstack.org/564564 | 20:15 |
| openstackgerrit | James E. Blair proposed openstack-infra/zuul master: Allow extra time for some ansible tests https://review.openstack.org/564572 | 20:15 |
| corvus | tobiash: that's with those 2 nits ^ | 20:15 |
| tobiash | corvus: added comment | 20:25 |
| mordred | corvus: oy. I have a bunch of reading to catch up on don't I? | 20:30 |
| corvus | i have a test locally with a really big (openstack-size) configuration; it shows a dynamic layout generation causing memory to increase by about 150MB, with the new patchset, it increases by about 0.5MB. | 20:30 |
| corvus | so i think we might be heading in the right direction :) | 20:31 |
| tobiash | Yay that sounds awesome :) | 20:32 |
| corvus | tobiash: whoops, i'll fix that up real quick :) | 20:32 |
| openstackgerrit | James E. Blair proposed openstack-infra/zuul master: Cache configuration objects in addition to YAML dicts https://review.openstack.org/564061 | 20:33 |
| openstackgerrit | James E. Blair proposed openstack-infra/zuul master: Stop deep-copying job variables https://review.openstack.org/564564 | 20:33 |
| openstackgerrit | James E. Blair proposed openstack-infra/zuul master: Allow extra time for some ansible tests https://review.openstack.org/564572 | 20:33 |
| corvus | tobiash: ^ | 20:33 |
| corvus | mordred: and yes, reading, but it's fun! the configloader stuff is becoming incrementally easier to work with. and if we can push up 50 zuul.yaml changes it will be a happy day | 20:34 |
| tobiash | corvus: now you get +2 real quick :) | 20:35 |
| corvus | like, there's actually a method in configloader now that says "load things into cache. then parse them. then cache the parsed objects. then add the parsed objects to a layout". i mean, i can almost understand that. :) | 20:35 |
| clarkb | corvus: its still on my todo list to review these things but my brother just decided to buy a boat so I'm getting roped into that :/ | 20:35 |
| corvus | clarkb: i'm good with boats; you review, i'll deal with the boat. | 20:36 |
| clarkb | ha | 20:36 |
| corvus | see its all about teamwork in open source projects | 20:36 |
| *** yolanda__ has quit IRC | 20:38 | |
| tobiash | corvus: yes, I was surprised that it's much easier now | 20:38 |
| mordred | clarkb: what kind of boat? | 20:38 |
| clarkb | mordred: a little drift boat | 20:39 |
| corvus | (also, you need to coil the ropes neatly so you *don't* get roped into the boat) | 20:39 |
| *** sshnaidm is now known as sshnaidm|off | 20:39 | |
| clarkb | for fiahing on small rivers | 20:39 |
| tobiash | corvus: today I got management approval for vancouver so I'll be there :) | 20:41 |
| corvus | tobiash: yay! | 20:41 |
| *** yolanda__ has joined #zuul | 20:43 | |
| mordred | \o/ | 21:02 |
| clarkb | ya I noticed you had signed up for the dinner thing \o/ | 21:12 |
| mordred | clarkb: oh - crap. did I sign up for the dinner thing? | 21:15 |
| clarkb | mordred: https://ethercalc.openstack.org/7vm2xrsk1yju not yet, currently just taking names and availability | 21:16 |
| clarkb | I sent it to the infra list. If zuulians will be there too feel free to join us | 21:16 |
| clarkb | Monday looks best which is apparently a holdiay which may make some of this tricky but whatever | 21:16 |
| openstackgerrit | James E. Blair proposed openstack-infra/zuul master: Report git sha in status page version https://review.openstack.org/564628 | 21:24 |
| openstackgerrit | James E. Blair proposed openstack-infra/zuul master: Report git sha in status page version https://review.openstack.org/564628 | 21:39 |
| SpamapS | hrm.. | 21:43 |
| SpamapS | I have a post job that is trying to sync logs from target->executor and with one particular image it gets a perm denied... :-/ | 21:44 |
| SpamapS | And zuul's key doesn't seem to be on the box | 21:44 |
| SpamapS | so a held node can't be accessed | 21:44 |
| *** hashar has quit IRC | 21:44 | |
| SpamapS | I wonder if I accidentally removed it during the job | 21:44 |
| clarkb | SpamapS: we use noav metadata to add our root users' keys to the root user | 21:49 |
| clarkb | doesn't prevent things from removing those keys though | 21:49 |
| SpamapS | Yeah I don't do anything like that.. but zuul's key should be able to get in. | 21:52 |
| SpamapS | I think I broke that | 21:52 |
| SpamapS | it's the job, not the image | 21:52 |
| SpamapS | ah yep, found it | 21:53 |
| SpamapS | job says it is "adding" a key | 21:53 |
| SpamapS | but it sets the key to just that | 21:53 |
| SpamapS | and because of persistent ssh, that didn't break, until this new base post task I added to pull the journals down | 21:53 |
| *** dkranz has quit IRC | 21:56 | |
| * SpamapS renames this channel to #rubberducky | 21:58 | |
| *** ssbarnea_ has quit IRC | 22:11 | |
| *** rlandy is now known as rlandy|afk | 22:19 | |
| clarkb | corvus: comment on https://review.openstack.org/#/c/564563/1 | 22:31 |
| clarkb | corvus: I've +2'd it but wanted to make sure you saw that | 22:31 |
| corvus | clarkb: yeah, probably so | 22:36 |
| openstackgerrit | James E. Blair proposed openstack-infra/zuul master: Fix race in test_bubblewrap_leak https://review.openstack.org/564640 | 22:38 |
| corvus | clarkb: oh! i remember | 22:40 |
| corvus | clarkb: it's used by the constructor | 22:40 |
| corvus | wait, no it's used by the manager | 22:41 |
| corvus | yeah, so as long as it's set by the time the manager is added, it should be fine | 22:41 |
| openstackgerrit | James E. Blair proposed openstack-infra/zuul master: Remove source_context argument to Pipeline https://review.openstack.org/564642 | 22:42 |
| clarkb | corvus: comments on https://review.openstack.org/#/c/564061/6 I think it is functionally correct, just some things I noticed when doing a deep review of that change that could be improved from before and some improvements maybe for the new stuff | 23:22 |
| clarkb | ok I think I've reviewed that entire stack at this point. | 23:23 |
| clarkb | Seems worthy of a release all on its own | 23:23 |
| clarkb | but we should probably give that a good long while on openstack before doing so? | 23:23 |
| clarkb | anyways let me know if you have questions about my comments | 23:23 |
| corvus | clarkb: yeah, maybe we can restart it tomorrow | 23:24 |
| clarkb | fwiw I find this much easier to understand as well | 23:24 |
| corvus | clarkb: replied. i agree 3 of those are potential future improvements. i think the sourcecontext thing could go either way (update now or later) -- can you double check i understood your comment correctly on that one? | 23:26 |
| clarkb | ya looking | 23:27 |
| clarkb | corvus: responded in gerrit | 23:30 |
| corvus | clarkb: hrm, one thing i like about the current method is we're certain we're not getting any implied branch matchers or anything. i don't *think* that's an issue here, but i'll need to double check. | 23:30 |
| openstackgerrit | James E. Blair proposed openstack-infra/zuul master: Simplify UnparsedConfig.copy https://review.openstack.org/564647 | 23:31 |
| corvus | yeah, it should be fine because those only get set on the copied sourcecontext objects. | 23:33 |
| corvus | (so our deepcopied copy will have the default values of None) | 23:33 |
| *** rlandy|afk is now known as rlandy | 23:48 | |
| SpamapS | Has anyone gotten zuul to deploy itself? | 23:54 |
| SpamapS | I have been deploying from my prod zuul -> my stage zuul for the last few months | 23:55 |
| SpamapS | but I recently turned on deploying to prod | 23:55 |
| SpamapS | and it keeps dying randomly later in the job | 23:55 |
| SpamapS | I have no definitive evidence that this is a problem (the deploy scripts do not restart any services) | 23:56 |
| clarkb | SpamapS: well if you restart the executor it will kill the jobs running. And if you restart the scheduler it releases all the node locks and will kill the job | 23:56 |
| clarkb | oh if you don't restart services then I don't think it should be a problem | 23:56 |
| SpamapS | but like this last time zuul noticed the tenants config change we made, and the deploy job kinda just.. froze shortly thereafter. | 23:56 |
| SpamapS | Right I don't restart executor. | 23:57 |
| SpamapS | For that reason :) | 23:57 |
| SpamapS | also because I don't want to restart anything until I have an automated queue saver. | 23:57 |
| SpamapS | which I don't :-P | 23:57 |
| SpamapS | so I tend to just let the deploys go, and then restart when the queues are empty | 23:57 |
| clarkb | our deployment is basically that, it updates configs and installs the new python bits | 23:58 |
| clarkb | but no restarting and is fine | 23:58 |
| SpamapS | yeah same | 23:58 |
| SpamapS | though some configs cause things to happen, such as tenants.yaml | 23:58 |
| clarkb | its not via zuul though | 23:58 |
| SpamapS | Well like I said, I have no definitive proof | 23:58 |
| clarkb | SpamapS: ya but just config relaods, so unless it is crashing doing that I think it should be fine | 23:58 |
| SpamapS | it's also on CentOS 7 and I suspect SSH just has bugs. | 23:59 |
| clarkb | you probably want to look at your executor logs to see why the job is having the sad | 23:59 |
| SpamapS | clarkb: SSH timed out | 23:59 |
| clarkb | SpamapS: I would be suspicious of actualyl restarting zookeeper or the scheduler in that case | 23:59 |
| SpamapS | 2018-04-26 16:35:04,658 DEBUG zuul.AnsibleJob: [build: 807c2b66a8894b6b8f770a164a58d16e] Ansible output: b'fatal: [godaddy-c7]: UNREACHABLE! => {"changed": false, "msg": "SSH Error: data could not be sent to | 23:59 |
| SpamapS | remote host \\"10.26.57.123\\". Make sure this host can be reached over ssh", "unreachable": true}' | 23:59 |
| clarkb | since that will cause the lock to go away and nodepool will delete the nodes | 23:59 |
Generated by irclog2html.py 2.15.3 by Marius Gedminas - find it at mg.pov.lt!