AWX Office Hours Agenda - Feb 14 2023

[john-westcott-iv]

AWX Office Hours

Proposed agenda based on topics

• @erz4: liveness and readiness prob for awx containers Add liveness/readiness probes to web/task - fixes #414  awx-operator#1188
• @stanislav-zaprudskiy graceful termination of awx pods AWX: Add termination_grace_period_seconds awx-operator#1193 and Add disable_instance management command #13445
• @TheRealHaoLiu receptor saga part 2: the horror of mismatched k8s API server and node version Job failed with just error and without log output #13469 and Do not use reconnect by default receptor#729
• @fosterseth tagging awx-ee images with the AWX release tag awx-ee latest on awx release #13338
• @relrod stream9 without ansible-builder (via @shanemcd) awx-ee#160 and Install missing dependencies awx-ee#165
• @miles-w-3 Include previous helm chart in helm index awx-operator#1183 fix to include previous version of helm chart in helm index
• @mani3887: Can we use redis to help resume job after awx controlplane pod restart
• @nickjmv: How to enable SAML via the extra_settings instead of the GUI.
• @TheRealHaoLiu first time contributor Spud light
  • Congratulations to https://github.com/OscarBell for his first contribution to the AWX project Fix verbosity parameter choices for ad_hoc_command module #13379
  • Congratulations to https://github.com/OscarBell for his first contribution to the AWX project Fix verbosity parameter choices for ad_hoc_command module #13379
• @TheRealHaoLiu announcing the recipient of the "Angry Spud" award
  • Congratulations to @philipsd6 for having the oldest open PR in the repo that @AlanCoding will eventually get to.

What

After a successful Contributor Summit in October, one of the bits of feedback we got was to host a regular time for the Automation Controller (AWX) Team to be available for your folks in the AWX Community, so we are happy to announce a new regular video meeting.

This kind of feedback loop is vital to the success of AWX and the AWX team wants to make it as easy as possible for you - our community - to get involved.

Where & When

Our next meeting will be held on Tuesday, Feb 14th, 2023 at 1500 UTC

• Google Meet
• Via Phone PIN: 842522378 Guide

This meeting is held once a month, on the second Tuesday of the month, at 1500 UTC

How

Add one topic per comment in this GitHub issue

If you don't have a GitHub account, jump on #awx:ansible.com on Matrix and we can add the topic for you

Talk with us

As well as the fortnightly video meeting you can join the Community (inc development team) on Matrix Chat.

• Matrix: #awx:ansible.com (recomended)
• libera.chat IRC: #ansible-awx (If you are already setup on IRC)

The Matrix & IRC channels are bridged, you'll just have a better experience on Matrix

Links

AWX YouTube Chanel
Previous Meeting
Meeting recording
Next Meeting

See you soon!

[john-westcott-iv]

From @mani3887:

Hi Folks,

Thanks for the AWX office hours. I went through the recording and other videos listed in Ansible AWX community. The videos were helpful in giving clarity of some of the designs followed in AWX. We are running AWX version 21.4 in AKS cluster. At our application we observe a below issue:

When AWX control pod restarts, it seems like it abandons the previous control pod's AWX EE pods, i.e. it seems like the new AWX control pod does not detect the already running AWX EE pods. Would it help if we move the redis cache from in-memory to cloud(azure redis cache)? Is there support for this in AWX? If there is, if we have multiple AWX replicas, and all having redis cache in cloud, would there be sync issues across the AWX instances? I understand that if we did move to a cloud-based solution from in-memory, the latency would be higher.
We had the below questions:

Redis cache - In the videos, it has been mentioned that in-memory redis is primarily used for dumping the logs, dispatcher gets the logs from EE pods, sends it to redis cache, the callback handler gets this from the redis cache and dumps it in the DB. Is the redis cache used for any other functionality? Is this only for the ansible job output logs or it also does the same for the metadata of the job? In our application, we are currently re-directing the Ansible logs(executed on EE pod) to a mounted file share; would the AWX still gather the logs and forward it to dispatcher? we have a solution for transferring the logs, but ideally we would want the metadata for the logs as well.
Role of AWX task container especially during the spinning up of the new pods; does it reconcile with the DB and then spin up the pod; while viewing the youtube videos, it seemed like when we hit the launch button on the UI, a pod would move to Pending state, my question is: would this information also be stored in the DB?
Thanks in advance.
Regards,
Mani

[john-westcott-iv]

From @nickjmv:

A tutorial on how to enable SAML via the extra_settings instead of the GUI.

This would (probably) be a solution for this: #12238

[erz4]

We want to advance that topic (pr) we encounter on issues when the deployment is rolled out
ansible/awx-operator#1188

[TheRealHaoLiu]

congratulations to https://github.com/mahaffey for his first contribution to the AWX project #13483

[TheRealHaoLiu]

congratulations to https://github.com/OscarBell for his first contribution to the AWX project #13379

[stanislav-zaprudskiy]

Let's discuss ansible/awx-operator#1193 (and #13445) during the meeting

[TheRealHaoLiu]

receptor saga pt2 - the horror of mismatched k8s API server and node version

#13469

ansible/receptor#729

TL;DR: RECEPTOR_KUBE_SUPPORT_RECONNECT now disable by default

please only re-enable after checking node version

[miles-w-3]

Topic:
Walk through resolution of helm index -

• Build index file each release and push it to gh-pages branch
• Now just use helm-index instead of cr
  Discuss limitations of helm index -
• solution is built from scratch instead of using wide
  • necessary because the fully-built chart is not checked into git, so the cr pipeline does not detect differences on each release
• helm-index overhead
  • requires downloading all previous releases in order to hash them and put them in the index file
  • dates are not accurate in the index file, since they generate
• Potential improvement
  • use yq or something similar to see which releases are already in index yaml, only add new release
  • This would lessen overhead, and not increment all dates on each release
  • This would leverage the merge flag in helm index which currently isn't used, since the command regenerates and overwrites conflicting entries in the yaml file

[TheRealHaoLiu]

@relrod talk about new awx-ee stream9 base image

[AlanCoding]

A candidate patch for ansible/ansible-runner#1187 is up, and that will affect a good number of people who hit this kind of error.