On Tue, Apr 15, 2014 at 2:52 AM, Kyotaro HORIGUCHI
<horiguchi.kyotaro@lab.ntt.co.jp> wrote:
> Hello, thank you for the discussion.
>
> At Tue, 1 Apr 2014 11:41:20 -0400, Robert Haas wrote
>>> I don't find that very radical at all. The backup_label file is
>>> *supposed* to be removed on the master if it crashes during the
>>> backup; and it should never be removed from the backup itself. At
>>> least that's how I understand it. Unfortunately, people too often
>
> The code indeed seems to assume that, and I couldn't think of any
> measure to avoid that dead-end once recovery sequence reads
> backup label accidentially left behind. I thought up to remove
> backup label during immediate shutdown on prvious versions, like
> 9.4 does.
>
> CancelBackup does only stat-unlink-rename sequence so I think
> this doesn't obstruct immediate shutdown sequence. And this
> doesn't change any seeming behavior or interfaces just except for
> this case. What do you think about this? Isn't this also
> applicable for older versions?
I don't think we should consider changing long-established behavior in
the back-branches. The old behavior may not be ideal, but that
doesn't mean it's a bug.
--
Robert Haas
EnterpriseDB: http://www.enterprisedb.com
The Enterprise PostgreSQL Company