Details
-
Admin & Maintenance Task
-
Status: Done
-
Medium
-
Resolution: Done
-
2.26.0
-
1
-
Yes
-
Yes
-
[obsolete] Architecture & Foundation
Description
Problem:
After we migrated from Prometheus to VictoriaMetrics, we never attempted to remove prometheus.ini config for supervisor. If the User sequentially runs `supervisorctl stop all` and `supervisorctl start all`, they will notice that supervisor tried and failed to start Prometheus.
% supervisorctl start all ... prometheus: ERROR (spawn error)
Later, when they try to print supervisord's status, they will see that supervisor has been relentlessly trying to launch prometheus' daemon, but all attempts failed, since on PMM versions starting from 2.12.0 `prometheus` is not installed.
% supervisorctl status ... prometheus BACKOFF Exited too quickly (process log may have details) # a bit later % supervisorctl status prometheus FATAL Exited too quickly (process log may have details)
This is definitely redundant and the User would benefit if we remove this noice from the system.
Solution:
Remove prometheus.ini it from PMM.
Suggested implementation:
- Leverage ansible to remove the package named percona-prometheus, which may have been kept on versions prior to 2.12.0
- Remove the generation of `prometheus.ini` config from `pmm-managed`, which is responsible for it now.
How to test:
- after executing `docker exec -it pmm-server sh -c "supervisorctl status"` against a running pmm-server, you should get:
% docker exec -it pmm-server sh -c "supervisorctl status"
alertmanager RUNNING pid 280, uptime 0:00:29
clickhouse RUNNING pid 64, uptime 0:00:36
dbaas-controller RUNNING pid 326, uptime 0:00:28
grafana RUNNING pid 68, uptime 0:00:36
nginx RUNNING pid 69, uptime 0:00:35
pmm-agent RUNNING pid 101, uptime 0:00:35
pmm-managed RUNNING pid 92, uptime 0:00:35
pmm-update-perform STOPPED Not started
pmm-update-perform-init EXITED Aug 06 01:19 AM
postgresql RUNNING pid 63, uptime 0:00:36
qan-api2 RUNNING pid 237, uptime 0:00:32
victoriametrics RUNNING pid 73, uptime 0:00:35
vmalert RUNNING pid 75, uptime 0:00:35
- be sure that `prometheus` is not listed
- be sure that `prometheus.ini` is absent in `/etc/supervisord.d/`
