Re: Harry Ambrose 2017-08-25 
<cak4knu86q98fwnzyntszk1ugca6f129wlshmrkzyb2ezksm...@mail.gmail.com>
> Hi All,
> 
> Sorry to open this can of worms again. However, we are still struggling
> with this issue across quite a large amount of our estate.

Hi,

we've just seen exactly this error on a customer database running
9.5.3 (postgresql95.x86_64 9.5.3-2PGDG.rhel6). Luckily just one tuple
was affected.

Symptoms were:
# select text from k... where id = 719764749;
ERROR:  unexpected chunk number 0 (expected 1) for toast value 3347468184 in 
pg_toast_922511637
The toast table itself was perfectly ok, with a single chunk:
# select * from pg_toast.pg_toast_922511637 where chunk_id = 3347468184;
chunk_id   | 3347468184
chunk_seq  | 0
chunk_data | ...valid text string... (with bytea_output = 'escape')

Updating or deleting the field/row didn't work:
# update k... set text = '...same text as above...' where id = 719764749;
ERROR:  XX000: tuple concurrently updated
ORT:  simple_heap_delete, heapam.c:3171

# delete from k... where id = 719764749;
ERROR:  XX000: tuple concurrently updated
ORT:  simple_heap_delete, heapam.c:3171

The problem persisted over the last two weeks (eventually noticed by
pg_dump starting to fail between August 15th and 18th). The server was
started on July 26th.

Besides the ERRORing statements above, I didn't actively resolve it,
suddenly SELECTing the original row just worked again. According to
pg_stat_user_tables, autovacuum didn't hit in. I can't say if there
were backends open for two weeks. At the time it resolved itself, the
oldest backend was from August 27th.

If xmin/xmax/multixact... data from this server is interesting, I can
extract it on request.

Christoph


-- 
Sent via pgsql-general mailing list (pgsql-general@postgresql.org)
To make changes to your subscription:
http://www.postgresql.org/mailpref/pgsql-general

Reply via email to