Hello Tom, thanks for your answer!
We found out because we are monitoring long running queries, and
saw it had been running for a month before the restart yesterday.
I just queried pg_stat_activity and it seems to be running since
then.
taimusz=# SELECT pid, query_start, usename, left(query,70)
FROM pg_stat_activity
WHERE query != '<IDLE>' AND query NOT ILIKE
'%pg_stat_activity%'
ORDER BY query_start;
pid | query_start | usename
| left
---------+-------------------------------+------------+------------------------------------------------------------------------
2321161 | 2021-05-17 16:15:13.906679+02 | subscriber | SELECT
pg_catalog.set_config('search_path', '', false);
SELECT pg_catalog.set_config('search_path', '', false);What makes you think it "keeps running"? It looks to me like the replication client does that and then goes about its business:
Admittedly, since you seem to have omitted the PID from your log_line_prefix, it's hard to be 100% sure that these log entries are from the same process. But I bet they are. The standard walreceiver definitely does things this way.
Sorry for omitting the PIDs. But you are probably right, it is a test environment where not much is happening
In short, I think there's nothing to see here. regards, tom lane
I wish this is the case. There seem to be no bloated tables, but before trying this on a production system I want to be sure
regards
Hannes
 |
|
