Exclude pg_dump sessions from PostgresqlSlowQueries alert

[souravbiswassanto]

Problem

PostgresqlSlowQueries fires on every pg_dump backup:

PostgreSQL has a transaction running longer than 60s on ace-db-2.
  VALUE = 136.308345
  LABELS = map[__name__:pg_stat_activity_max_tx_duration application_name:pg_dump backend_type:client backend container:exporter datname:ace endpoint:metrics instance:10.2.4.8:56790 job:ace-db-stats namespace:ace pod:ace-db-2 server:localhost:5432 service:ace-db-stats state:active usename:postgres]

pg_dump holds a single repeatable-read transaction open for the entire duration of the dump to get a consistent snapshot, so pg_stat_activity_max_tx_duration for that session grows to however long the backup takes. That is expected behaviour for a backup, not a slow query, so the alert is pure noise on any database large enough for a dump to exceed the threshold.

Change

Add application_name!~"pg_dump.*" to the alert expression in charts/postgres-alerts/templates/alert.yaml, which drops both pg_dump and pg_dumpall sessions. Client workloads that are genuinely slow still trigger the alert unchanged.

Rendered output:

- alert: PostgresqlSlowQueries
  expr: pg_stat_activity_max_tx_duration{job="test-stats",namespace="ace",state="active",application_name!~"pg_dump.*"} > 60

Verified with helm template.

---

^{Need help on this PR? Tag @codesmith-bot with what you need. Autofix is disabled.}