[pgpool-general: 8563] Re: Issues taking a node out of a cluster

[Tatsuo Ishii]

> We've just completed a full test run where we use a stable numbering
> (possibly with gaps) for backends and a contiguous numbering for the
> watchdogs and all tests pass. This does have the disadvantage that a
> watchdog may get a different number in the configuration than the backend
> running on the same node, but this is mostly cosmetically. This solves this
> issue for us.

Great.

> The other question does still remain open though: Why does the
> auto_failback not reattach backend 1 when it detects the database is up and
> streaming? Could this have been related to the inconsistent numbering of
> the backends?

Other developer suggested that health check process for backend 1 may
not run in this case. (in order to make auto_failback working, health
check process for the backend needs to run). Can you confirm?

> I was under the impression that pgpool should always reattach
> a detached  backend when it is streaming from the primary.

Best reagards,
--
Tatsuo Ishii
SRA OSS LLC
English: http://www.sraoss.co.jp/index_en/
Japanese:http://www.sraoss.co.jp