Re: postgresql.auto.conf read from wrong directory

From: Fujii Masao <masao(dot)fujii(at)gmail(dot)com>
To: Amit Kapila <amit(dot)kapila16(at)gmail(dot)com>
Cc: Christoph Berg <cb(at)df7cb(dot)de>, Andres Freund <andres(at)2ndquadrant(dot)com>, PostgreSQL Hackers <pgsql-hackers(at)postgresql(dot)org>
Subject: Re: postgresql.auto.conf read from wrong directory
Date: 2014-05-10 23:08:16
Message-ID: CAHGQGwH74ZC7ava5502MCZ2jAvPM3WGUQA3auHLf5xcbsAHrHw@mail.gmail.com
Views: Raw Message | Whole Thread | Download mbox | Resend email
Thread:
Lists: pgsql-hackers

On Sat, May 10, 2014 at 12:30 PM, Amit Kapila <amit(dot)kapila16(at)gmail(dot)com> wrote:
> On Fri, May 9, 2014 at 7:01 PM, Fujii Masao <masao(dot)fujii(at)gmail(dot)com> wrote:
>> On Fri, May 9, 2014 at 1:06 PM, Amit Kapila <amit(dot)kapila16(at)gmail(dot)com> wrote:
>>> There is no harm in forbidding data_directory, but similarly we can
>>> imagine that people can set some very large values for some config
>>> variables due to which later it can have symptoms similar to this
>>> issue.
>>
>> Yes, that can prevent the server from restarting at all. In this case,
>> to restart the server, we need to edit postgresql.auto.conf manually
>> and remove the problematic settings though the header of
>> postgresql.auto.conf warns "Do not edit this file manually!".
>
> So lets not try to forbid data_directory in postgresql.auto.conf and just
> fix the case reported in this mail for postgresql.conf for which the
> patch is attached upthread.

ISTM that data_directory is in different situation from the other. That is,
setting data_directory in postgresql.auto.conf is problematic whether its
setting value is valid or invalid. Imagine the case where data_directory
is set to '/data1' and '/data2' in config_directory/postgresql.conf and
/data1/postgresql.auto.conf, respectively. In this case, firstly the server
doesn't read /data2/postgresql.auto.conf when it starts up, even if your
patch has been applied. Secondly, the server doesn't read
/data1/postgresql.auto.conf when it receives SIGHUP, and this causes
the following error.

LOG: parameter "data_directory" cannot be changed without restarting the server

Regards,

--
Fujii Masao

In response to

Responses

Browse pgsql-hackers by date

  From Date Subject
Next Message Tom Lane 2014-05-10 23:08:48 Re: New pg_lsn type doesn't have hash/btree opclasses
Previous Message Michael Paquier 2014-05-10 22:45:20 Re: New pg_lsn type doesn't have hash/btree opclasses