| From: | surya poondla <suryapoondla4(at)gmail(dot)com> |
|---|---|
| To: | Bharath Rupireddy <bharath(dot)rupireddyforpostgres(at)gmail(dot)com> |
| Cc: | Sami Imseih <samimseih(dot)pg(at)gmail(dot)com>, Scott Ray <scott(at)scottray(dot)io>, PostgreSQL Hackers <pgsql-hackers(at)lists(dot)postgresql(dot)org> |
| Subject: | Re: pg_xmin_horizon: a system view of everything pinning the xmin horizon |
| Date: | 2026-11-17 21:32:33 |
| Message-ID: | CAOVWO5pMMDgP9Nk2RiiVAhy9XySr7k5Ns9pWr7-Bv-pkQS-CgQ@mail.gmail.com |
| Views: | Whole Thread | Raw Message | Download mbox | Resend email |
| Thread: | |
| Lists: | pgsql-hackers |
Hi Bharath, Scott,
Bharath's v11 removes the duplicated classifier by recording the inputs
inside ComputeXidHorizonsInternal(), which addresses the concern
Horiguchi, Sami and Bharath raised previously, and an xid8 xmin makes the
obvious
ORDER BY work.
One thing I'd like to keep from v10 is the xid column. In v11 a backend
row's xmin is the older of its xid and xmin, so a backend that owns the
oldest xid and backends that merely took snapshots after it look the
same. With one idle-in-transaction session and twenty readers whose
xmin equals its xid, that's twenty-one identical rows, and only one of
them is worth terminating. Scott added xid in v5 for that reason. The
proc's xid is still available at the point where v11 records its inputs,
so keeping it would cost one field in XidHorizonProc and one column.
Regards,
Surya Poondla
| From | Date | Subject | |
|---|---|---|---|
| Next Message | Manu | 2026-11-17 21:40:47 | Re: Assertion failure with temporary relations in single-user mode |
| Previous Message | Tomas Vondra | 2026-11-17 20:42:06 | Re: Assert failure in parallel query due to SubPlan double visit for hash join |