[citest]

[citest]

@natoscott The latest sync of code from ansible-pcp brings tests/tests_verify_pmie_webhook.yml and check_pmie_webhook.yml which seems very similar to the new proposed test tests_verify_notification.yml and check_notification.yml - is there some overlap? Should we get rid of one or the other?

@natoscott The latest sync of code from ansible-pcp brings tests/tests_verify_pmie_webhook.yml and check_pmie_webhook.yml which seems very similar to the new proposed test tests_verify_notification.yml and check_notification.yml - is there some overlap? Should we get rid of one or the other?

I think we need both as ansible-pcp is usable without the metrics role, but our QE folk will want to verify with the metrics role as well.

[citest]

[citest]

@natoscott this is looking better - but still have the following problems

• cannot restart pmlogger on centos7
• problems with redis on fedora
• problems with grafana on rhel 9

Can you take a look at these? They are unrelated to this PR, probably caused by something in that last sync with ansible-pcp

Can you take a look at these? They are unrelated to this PR, probably caused by something in that last sync with ansible-pcp

Sure thing, thanks for the heads-up @richm - will let you know what I find.

@richm OK, have discovered a few things. There are three failures causing all CI issues:

1. on CentOS-7 there's a strange pmlogger service failure - this is not a known issue, nothing has changed in el7 PCP for a long time - I wonder if this might be selinux or firewall result (external factors). Can we extract the journalctl output from the command indicated in the failure message, from a CI system?
2. on Fedora-37 and Fedora-38 the new pmie-webhook test is failing with "Could not find the requested service redis: host" - Redis isn't used here though. So, I think this is actually passing, its a test failure that seems to be related to the way we interrogate service state. It might be fixed by the patch below, WDYT?
3. all RHEL-9.4 based test cases are failing with the grafana-server core you observed. This is not a known issue and I'm chasing it down further with QE and Grafana devs in the team ... still WIP. There was a recent change to add selinux policy to Grafana in RHEL, so I'm wondering if this could be involved. Will get back to you tomorrow hopefully on this one.

tests/restore_services_state.yml
--- /tmp/git-blob-AYZrfz/restore_services_state.yml     2023-12-11 14:04:12.002996635 +1100
+++ tests/restore_services_state.yml    2023-12-11 13:57:08.143243811 +1100
@@ -16,6 +16,7 @@
   when:
     - item + '.service' in final_state.ansible_facts.services
     - item + '.service' in initial_state.ansible_facts.services
+    - initial_state.ansible_facts.services[item + '.service']['status'] != 'not-found'
   with_items:
     - pmcd
     - pmlogger

@richm have confirmed with our Grafana gurus that an SE Linux AVC has been observed to cause a Grafana core dump just recently. There is also a fixed grafana-selinux policy build recently submitted in 9.4 that may resolve this issue - if not, its quite likely to be a new selinux issue we've not seen before.

closing in favor of #183