Reading supplementary groups is too slow

[Andrew Pollock]

Binary package hint: postgresql-common

I'm using the Postgresql packages in an environment having 6,000+ local groups.
In this situation, the change_uid function in PgCommon.pm is unbearably slow. e.g.:

apollock@procrastination:~$ sudo time perl -I /usr/share/postgresql-common/ -MPgCommon -e 'change_ugid(29,29)'

4540.63user 426.61system 1:22:49elapsed 99%CPU (0avgtext+0avgdata 0maxresident)k
0inputs+0outputs (0major+678714minor)pagefaults 0swaps

This has already been fixed in version 90 of postgresql-common, and fixed in Intrepid and later. I'd like this to be fixed in Hardy as well. I will provide a minimal patch against 87

SRU justification: Makes standard PostgreSQL operations, such as starting the cluster from the init script, unbearably slow in environments with many users/groups.
Patch: http://bazaar.launchpad.net/~pitti/postgresql/common/revision/873 (in production use for over half a year without apparent regressions).
Test case: see above

For verification, you should also run the test suite with
  sudo /usr/share/postgresql-common/testsuite
This provides a very comprehensive integration test (> 1000 test cases). If that fails for you (I might have some system specific assumptions in there), please compare with the results with the hardy final postgresql-common.

[Martin Pitt (pitti)]

The necessary change is

  http://bazaar.launchpad.net/~pitti/postgresql/common/revision/873

I'm ok with backporting this.

[Andrew Pollock (apollock)]

Terrific, thanks. Do you want a debdiff or are you good to do the whole thing?

[Martin Pitt (pitti)]

Andrew Pollock [2009-04-22 15:03 -0000]:
> Terrific, thanks. Do you want a debdiff or are you good to do the
> whole thing?

That's fine, I can just cherrypick the commit. However, I'd appreciate
if you could test the package in -proposed (once uploaded) to have a
four-eyes verification process.

[Andrew Pollock (apollock)]

Sure thing

[Martin Pitt (pitti)]

I uploaded the backport now, sorry for dragging this so long! This needs Steve or Colin to process now.

[Steve Langasek (vorlon)]

Accepted into hardy-proposed, the package will build now and be available in a few hours. Please test and give feedback here. See https://wiki.ubuntu.com/Testing/EnableProposed for documentation how to enable and use -proposed. Thank you in advance!

[Martin Pitt (pitti)]

Andrew, can you please give this some testing? Thanks in advance!

[Joel Ebel (jbebel)]

I've testing this package, and it does start very quickly compared to the current hardy version.

[Launchpad Janitor (janitor)]

This bug was fixed in the package postgresql-common - 87ubuntu1

---------------
postgresql-common (87ubuntu1) hardy-proposed; urgency=low

  * PgCommon.pm, change_ugid(): Use /usr/bin/id instead of manual parsing with
    getgrent(), which is reportedly magnitudes faster for environments with
    a large number of groups in remote databases. (LP: #364669, Debian #491136)
    Patch backported from trunk r873 (applied in Ubuntu 8.10 and later).

 -- Martin Pitt <email address hidden> Sun, 19 Jul 2009 15:13:15 +0200