Horton

Re: Intermittent buildfarm failures on wrasse

Tom Lane <tgl@sss.pgh.pa.us>

View thread View on postgr.es
From: Tom Lane <tgl@sss.pgh.pa.us>
To: Andres Freund <andres@anarazel.de>
Cc: pgsql-hackers@lists.postgresql.org, Noah Misch <noah@leadboat.com>, Peter Geoghegan <pg@bowt.ie>, David Rowley <dgrowleyml@gmail.com>
Date: 2022-04-15T17:05:29Z
Lists: pgsql-hackers 
I wrote:
> Um, this is the logical replication launcher, not the autovac launcher.
> Your observation that a sleep in get_database_list() reproduces it
> confirms that, and I don't entirely see why the timing of the LR launcher
> would have changed.

Oh, to clarify: I misread "get_database_list()" as
"get_subscription_list()", which is the part of the LR launcher startup
that causes the problem for me.  So what we actually have confirmed is
that BOTH of those launchers are problematic for this.  And AFAICS
neither of them needs to be causing horizon adjustments for non-shared
tables.

(It's possible that the AV launcher is responsible in some of wrasse's
reports, but it's been the LR launcher in five out of five
sufficiently-instrumented failures for me.)

			regards, tom lane

Commits

  1. Repurpose PROC_COPYABLE_FLAGS as PROC_XMIN_FLAGS

  2. Tighten ComputeXidHorizons' handling of walsenders.

  3. Adjust VACUUM's removable cutoff log message.

  4. Temporarily add some probes of tenk1's relallvisible in create_index.sql.

  5. Set synchronous_commit=on in test_setup.sql.

  6. Rearrange core regression tests to reduce cross-script dependencies.

  7. snapshot scalability: Don't compute global horizons while building snapshots.