Merge branch 'infrastucture8494' into 'master'

Adding Vacuum Queue metrics to postgres-exporter.yaml

See merge request gitlab-org/omnibus-gitlab!3771

changelogs/unreleased/infrastucture8494.yml

+---
+title: Adding Vacuum Queue metrics to postgres-exporter.yaml
+merge_request: 3771
+author:
+type: added

files/gitlab-cookbooks/monitoring/templates/postgres-queries.yaml

@@ -172,3 +172,120 @@ pg_stuck_idle_in_transaction:
     - queries:
         usage: "GAUGE"
         description: "Current number of queries that are stuck being idle in transactions"
+
+pg_vacuum_queue:
+  #master: true 
+  # Until postgres_exporter is upgraded. See 
+  # https://gitlab.com/gitlab-org/omnibus-gitlab/issues/4887
+  query: |
+    with table_opts_vs_statistic as (
+      select
+        pg_class.oid,
+        pg_class.relname,
+        coalesce(nspname, 'public') as schemaname,
+        pg_class.relpages,
+        pg_class.reltuples,
+        case
+          when array_to_string(reloptions, '') like '%autovacuum_vacuum_threshold%' then regexp_replace(array_to_string(reloptions, ''), '.*autovacuum_vacuum_threshold=([0-9.]+).*', E'\\1')::int8
+          else current_setting('autovacuum_vacuum_threshold')::int8
+        end as autovacuum_vacuum_threshold,
+        case
+          when array_to_string(reloptions, '') like '%autovacuum_vacuum_scale_factor%' then regexp_replace(array_to_string(reloptions, ''), '.*autovacuum_vacuum_scale_factor=([0-9.]+).*', E'\\1')::numeric
+          else current_setting('autovacuum_vacuum_scale_factor')::numeric
+        end as autovacuum_vacuum_scale_factor,
+        case when array_to_string(reloptions, '') ~ 'autovacuum_enabled=(false|off)' then false else true end as autovacuum_enabled,
+        n_dead_tup,
+        last_autovacuum,
+        last_vacuum
+      from pg_class
+      join pg_namespace ns on relnamespace = ns.oid 
+      join pg_stat_all_tables psat  on psat.relid  = pg_class.oid
+      where relkind in ('r','m') 
+    ), p as (
+      select pgspv.*,a.query,a.wait_event_type,a.wait_event,a.query_start
+      from pg_stat_progress_vacuum pgspv
+      left join pg_stat_activity a using (pid)
+    ) 
+    select
+      table_opts_vs_statistic.schemaname as schemaname,
+      table_opts_vs_statistic.relname as relname,
+      round((100 * table_opts_vs_statistic.n_dead_tup::numeric / nullif(table_opts_vs_statistic.reltuples, 0))::numeric, 2) as dead_tup_pct,
+      table_opts_vs_statistic.reltuples::numeric as reltuples,
+      table_opts_vs_statistic.n_dead_tup,
+      (relpages::bigint*8*1024) AS table_size_bytes,
+      'V. Threshold:' || table_opts_vs_statistic.autovacuum_vacuum_threshold
+        || ', V. Scale Factor: ' || (table_opts_vs_statistic.autovacuum_vacuum_scale_factor)::numeric *100 ||' %'
+        || case when not autovacuum_enabled then ', DISABLED' else ', enabled' end as "effective_settings",
+      case
+        when last_autovacuum > coalesce(last_vacuum, '0001-01-01') then last_autovacuum::timestamp(0)
+        when last_vacuum is not null then last_vacuum::timestamp(0)
+        else null
+      end as "last_vacuumed",
+      case
+        when last_autovacuum > coalesce(last_vacuum, '0001-01-01') then 'auto'
+        when last_vacuum is not null then 'manual'
+        else null
+      end as "type",
+      coalesce(p.phase, 'in queue') as status,
+      p.pid as pid,
+      coalesce  (p.query,'')as  action,
+      case when p.pid is null then null else coalesce(wait_event_type ||'.'|| wait_event, 'f') end as waiting,
+      round(100.0 * p.heap_blks_scanned / nullif(p.heap_blks_total, 0), 1) AS scanned_pct,
+      round(100.0 * p.heap_blks_vacuumed / nullif(p.heap_blks_total, 0), 1) AS vacuumed_pct,
+      extract ('epoch' from  now()-query_start) elapsed_time
+    from 
+    table_opts_vs_statistic
+    full outer join p on p.relid = table_opts_vs_statistic.oid and p.datname = current_database()
+    where
+      table_opts_vs_statistic.relpages >= 8
+      and  autovacuum_vacuum_threshold + (autovacuum_vacuum_scale_factor::numeric * table_opts_vs_statistic.reltuples) < table_opts_vs_statistic.n_dead_tup
+  metrics:
+    - schemaname:
+        usage: "LABEL"
+        description: "Table Schema"
+    - relname:
+        usage: "LABEL"
+        description: "Table name"
+    - dead_tup_pct:
+        usage: "GAUGE"
+        description: "Estimated dead tuples percent"
+    - reltuples:
+        usage: "GAUGE"
+        description: "Number of tuples in table"
+    - n_dead_tup:
+        usage: "GAUGE"
+        description: "Estimated number of dead tuples"
+    - table_size_bytes:
+        usage: "GAUGE"
+        description: "Estimated table size"
+    - effective_settings:
+        usage: "LABEL"
+        description: "Autovacuums settings"
+    - last_vacuumed:
+        usage: "GAUGE"
+        description: "Last time at which this table was vacuumed"
+    - type:
+        usage: "LABEL"
+        description: "Last vacuum type"
+    - status:
+        usage: "LABEL"
+        description: "Vacuum actual status"
+    - pid:
+        usage: "GAUGE"
+        description: "Vacuum process id"
+    - action:
+        usage: "LABEL"
+        description: "Type of vacuum executed"
+    - waiting:
+        usage: "LABEL"
+        description: "Vacuum queue status"
+    - scanned_pct:
+        usage: "GAUGE"
+        description: "Estimated rows scanned percent"
+    - vacuumed_pct:
+        usage: "GAUGE"
+        description: "Estimated vacuumed rows percent"
+    - elapsed_time:
+        usage: "GAUGE"
+        description: "Elapsed time vacuuming (in seconds)"
+