postgres init script fails uncleanly if cluster directory is missing

[Philip Hofstetter]

Binary package hint: postgresql-common

On this particular setup, the cluster home directory /var/lib/postgresql is mounted to its own partition.

By accident, the directory wasn't mounted when I tried to launch postgres. Instead of a precise error message notifying me about the problem, I got a heap of perl error messages. While it's pure cosmetics, I think the scripts should check for basic errors like that and not bail out showing a bunch of perl errors.

Steps to reproduce:
- Create a postgres-cluster
- Remove the parent directory of said cluster (or unmount its partition)
- Try to launch postgres.

Expected tesults:
- Error-Message explaining the problem

Actual results:
root@vmmepha:~# /etc/init.d/postgresql-8.3 start
 * Starting PostgreSQL 8.3 database server
 * Use of uninitialized value in getpwuid at /usr/bin/pg_ctlcluster line 433.
Use of uninitialized value in getgrgid at /usr/bin/pg_ctlcluster line 436.
Use of uninitialized value in concatenation (.) or string at /usr/share/postgresql-common/PgCommon.pm line 649.
Use of uninitialized value in numeric eq (==) at /usr/share/postgresql-common/PgCommon.pm line 658.
Use of uninitialized value in numeric eq (==) at /usr/share/postgresql-common/PgCommon.pm line 658.
Use of uninitialized value in numeric eq (==) at /usr/share/postgresql-common/PgCommon.pm line 658.
Use of uninitialized value in numeric eq (==) at /usr/share/postgresql-common/PgCommon.pm line 658.
Use of uninitialized value in numeric eq (==) at /usr/share/postgresql-common/PgCommon.pm line 658.
Use of uninitialized value in numeric eq (==) at /usr/share/postgresql-common/PgCommon.pm line 658.
Use of uninitialized value in numeric eq (==) at /usr/share/postgresql-common/PgCommon.pm line 658.
Use of uninitialized value in numeric eq (==) at /usr/share/postgresql-common/PgCommon.pm line 658.
Use of uninitialized value in numeric eq (==) at /usr/share/postgresql-common/PgCommon.pm line 658.
Use of uninitialized value in numeric eq (==) at /usr/share/postgresql-common/PgCommon.pm line 658.
Use of uninitialized value in numeric eq (==) at /usr/share/postgresql-common/PgCommon.pm line 658.
Use of uninitialized value in numeric eq (==) at /usr/share/postgresql-common/PgCommon.pm line 658.
Use of uninitialized value in scalar assignment at /usr/share/postgresql-common/PgCommon.pm line 666.
Use of uninitialized value in scalar assignment at /usr/share/postgresql-common/PgCommon.pm line 667.
Use of uninitialized value in numeric ne (!=) at /usr/share/postgresql-common/PgCommon.pm line 668.
Use of uninitialized value in numeric ne (!=) at /usr/share/postgresql-common/PgCommon.pm line 669.
Use of uninitialized value in numeric ne (!=) at /usr/bin/pg_ctlcluster line 445.
pg_controldata: could not open file "/var/lib/postgresql/8.3/main/global/pg_control" for reading: No such file or directory
Error: Could not parse locale out of pg_controldata output
   ...fail!

$ lsb_release -rd
Description: Ubuntu hardy (development branch)
Release: 8.04

[Martin Pitt (pitti)]

IMHO, the "could not open file "/var/lib/postgresql/8.3/main/global/pg_control" for reading: No such file or directory" message is clear enough. It would be nice to suppress all the perl warnings, of course.

[Martin Pitt (pitti)]

Reproduced, and test case committed to postgresql-common development trunk.

[Martin Pitt (pitti)]

Fixed in development trunk, will upload to Debian soon.

[Launchpad Janitor (janitor)]

This bug was fixed in the package postgresql-common - 88

---------------
postgresql-common (88) unstable; urgency=low

  * t/030_errors.t: Add various test cases for stopping stopped clusters with
    stale or corrupt PID files, with and without --force. This reproduces
    #473879.
  * pg_ctlcluster: On stop, clean up stale/corrupt PID files if the server is
    not running. Also fix a taint error and simplify the code a bit.
    (Closes: #473879)
  * t/030_errors.t: Test graceful handling of absent /var/lib/postgresql.
    (Reproduces LP #210322). Adapt 020_create_sql_remove.t accordingly.
  * pg_ctlcluster: Check early whether the data directory exists and is
    accessible, to avoid lots of Perl warning clutter. (LP: #210322)
  * pg_createcluster POD: Clarify what "integrating existing cluster data
    directory" means: configuration files must already be present, we cannot
    create them out of thin air. (Closes: #475954)
  * pg_maintenance: Sort versions for predictable output (looks nicer and
    unbreaks the test suite).

 -- Ubuntu Archive Auto-Sync <email address hidden> Fri, 02 May 2008 02:23:00 +0100