valgog <valgog@gmail.com> writes:
>> This process seems almost entirely unrelated to the documented way of
>> doing it; I'm not surprised that you end up with some files not in sync.
>> Please see pg_start_backup and friends.
> It was done as documented in http://www.postgresql.org/docs/9.0/static/backup-incremental-updated.html
Oh, I missed that you were copying from a hot-standby slave rather than
the master. Still, your procedure doesn't clearly match step 2, and
that step is the weak point of the process --- the grandchild slave
isn't consistent until it's replayed WAL far enough, but we don't have
any automated support for verifying that. (I hope that's going to get
improved in 9.1.) I suspect you allowed the grandchild to go live
before it was really consistent.
regards, tom lane