Hi,
We are asked to have key monitoring or alerting added to our postgres database. And I am thinking of metrics like blocked transactions, Max used transaction Ids, Max Active session threshold, Deadlock, Long running query, replica lag, buffer cache hit ratio, read/write IOPS or latency etc. I have below questions
1)Below are some which i tried writing, can you please let me know if these are accurate?
2)How should we be writing the alerting query for deadlock, max used transaction ids, read/write IOPS and latency?
3)Are there any docs available which have these sample sql queries on the pg_* table for these critical alerts which we can easily configure through any tool?
We are asked to have key monitoring or alerting added to our postgres database. And I am thinking of metrics like blocked transactions, Max used transaction Ids, Max Active session threshold, Deadlock, Long running query, replica lag, buffer cache hit ratio, read/write IOPS or latency etc. I have below questions
1)Below are some which i tried writing, can you please let me know if these are accurate?
2)How should we be writing the alerting query for deadlock, max used transaction ids, read/write IOPS and latency?
3)Are there any docs available which have these sample sql queries on the pg_* table for these critical alerts which we can easily configure through any tool?
4)Any other alerts which we should be really having?
*****Blocking sessions
select distinct blocking_id from
(
SELECT
activity.pid,
activity.usename,
activity.query,
blocking.pid AS blocking_id,
blocking.query AS blocking_query
FROM pg_stat_activity AS activity
JOIN pg_stat_activity AS blocking
ON blocking.pid = ANY(pg_blocking_pids(activity.pid))
) a;
**** long running beyond ~1 hours*****
SELECT
query,
datname,
pid,
now() - state_change AS idle_for
FROM pg_stat_activity
WHERE state IN ('active', 'idle in transaction')
AND pid <> pg_backend_pid()
AND xact_start < now() - interval '1 hour'
ORDER BY age(backend_xmin) DESC NULLS LAST;
**** No of active sessions ******
SELECT count(*) AS active_connections
FROM pg_stat_activity
WHERE state = 'active';
***replica lag****
SELECT client_addr, state, sent_location, write_location, flush_location, replay_location,
pg_wal_lsn_diff(sent_location, replay_location) AS replica_lag
FROM pg_stat_replication;
***buffer cache hit ratio****
SELECT
(1 - (blks_read::float / (blks_hit + blks_read))) * 100 AS buffer_cache_hit_ratio
FROM pg_stat_database;
*****Blocking sessions
select distinct blocking_id from
(
SELECT
activity.pid,
activity.usename,
activity.query,
blocking.pid AS blocking_id,
blocking.query AS blocking_query
FROM pg_stat_activity AS activity
JOIN pg_stat_activity AS blocking
ON blocking.pid = ANY(pg_blocking_pids(activity.pid))
) a;
**** long running beyond ~1 hours*****
SELECT
query,
datname,
pid,
now() - state_change AS idle_for
FROM pg_stat_activity
WHERE state IN ('active', 'idle in transaction')
AND pid <> pg_backend_pid()
AND xact_start < now() - interval '1 hour'
ORDER BY age(backend_xmin) DESC NULLS LAST;
**** No of active sessions ******
SELECT count(*) AS active_connections
FROM pg_stat_activity
WHERE state = 'active';
***replica lag****
SELECT client_addr, state, sent_location, write_location, flush_location, replay_location,
pg_wal_lsn_diff(sent_location, replay_location) AS replica_lag
FROM pg_stat_replication;
***buffer cache hit ratio****
SELECT
(1 - (blks_read::float / (blks_hit + blks_read))) * 100 AS buffer_cache_hit_ratio
FROM pg_stat_database;
Regards
Yudhi
