Thanks much, I'll keep my eyes open today night hoping it will not happen again.
On Thu, Jul 19, 2018 at 5:39 PM, Tom Lane <tgl@sss.pgh.pa.us> wrote:
[ please keep the list cc'd for the archives' sake ] Alessandro Aste <alessandro.aste@gmail.com> writes: > Hello Tom, thanks for your reply: > SELECT * FROM pg_class WHERE OID = 2223152859 ; > (0 rows) > I'm not aware of any DDL at that time.
Hm. Well, that OID was definitely there when pg_dump looked, and it's not there now, so something changed --- though we can't prove it changed concurrently.
In any case, I'd bet that if we ran this to ground it would prove to be a concurrent-DDL issue. pg_dump tries to protect itself against concurrent DDL, but for assorted architectural reasons the protection is not 100%; sometimes you can get odd failures like this, essentially due to "clock skew" between pg_dump's view of the catalogs and the server's view of the catalogs. As long as it works on retry, I wouldn't worry too much about it.