[Pgpool-hackers] Fw: [HACKERS] Any idea for serializing INSERTING SERIAL column?

[Tatsuo Ishii]

> Hi, I have sent this question to pgsql-hackers list. This issue may be
> a showstopper for 3.0.4 release because with this we got a PostgreSQL
> error after XID wraparound happens. Returning back to table lock will
> arouse conflict with autovacuum problem happend in 2.x days.
> 
> Any thought?

After some discussions with Takahiro, I decided to release pgpool-II
3.0.4 soon without the fix for this.

The reasons for this are:

- Using advisory lock was suggested in pgsql-hackers. However using
  the method seems to make pgpool more fragile. Because it may
  introduce lock leaks.

- Adivisory lock is not available on all PostgreSQL releases (8.2 or
  later)

- Need carefull testing and will postpone the release at least 1 week
  or more.

In summary putting the fix in 3.0.4 is too leasky.

I will develop the fix (probably not using advisory lock) for 3.1 and
3.0.5 after 3.0.4 release.
--
Tatsuo Ishii
SRA OSS, Inc. Japan
English: http://www.sraoss.co.jp/index_en.php
Japanese: http://www.sraoss.co.jp