| From: | "Hayato Kuroda (Fujitsu)" <kuroda(dot)hayato(at)fujitsu(dot)com> |
|---|---|
| To: | 'Xuneng Zhou' <xunengzhou(at)gmail(dot)com> |
| Cc: | 赵宇鹏(宇彭) <zhaoyupeng(dot)zyp(at)alibaba-inc(dot)com>, pgsql-hackers <pgsql-hackers(at)lists(dot)postgresql(dot)org> |
| Subject: | RE: memory leak in logical WAL sender with pgoutput's cachectx |
| Date: | 2025-08-14 12:49:23 |
| Message-ID: | OSCPR01MB14966A4407EDF6111B5E49485F535A@OSCPR01MB14966.jpnprd01.prod.outlook.com |
| Views: | Whole Thread | Raw Message | Download mbox | Resend email |
| Thread: | |
| Lists: | pgsql-hackers |
Dear Xuneng,
> Is it safe to free the substructure from within rel_sync_cache_relation_cb()?
You referred the comment in rel_sync_cache_relation_cb() right? I understood like
that we must not access to any *system caches*, from the comment. Here we do not
re-build caches so that we do not access to the syscaches - it is permitted.
I'm happy if you also confirm the point.
> I’ also interested in the reasoning behind setting
> NINVALIDATION_THRESHOLD to 100.
This is the debatable point of this implementation. I set to 100 because it was
sufficient with the provided workload, but this may cause the replication lag.
We may have to consider a benchmark workload, measure data, and consider the
appropriate value. 100 is just an initial point.
Best regards,
Hayato Kuroda
FUJITSU LIMITED
| From | Date | Subject | |
|---|---|---|---|
| Next Message | Heikki Linnakangas | 2025-08-14 13:10:32 | Re: BackendKeyData is mandatory? |
| Previous Message | Ashutosh Bapat | 2025-08-14 12:44:29 | Re: Improve pg_sync_replication_slots() to wait for primary to advance |