Re: pg_dump crashing - Mailing list pgsql-general

On 03/20/2016 08:24 AM, Matthias Schmitt wrote:
> Hello,
>
>> On 16 Mar 2016, at 14:55, Adrian Klaver <adrian.klaver@aklaver.com> wrote:
>>
>> On 03/15/2016 08:10 AM, Matthias Schmitt wrote:
>>> Hello,
>>>
>>> since two weeks I am trying to get PostgreSQL 9.5.1 to run on Debian 8.3. Everything is fine except the daily
backup.When calling pg_dump as part of a cron job pg_dump crashes: 
>>>
>>> 2016-03-15 01:00:02 CETFATAL:  semctl(23232524, 3, SETVAL, 0) failed: Invalid argument
>>> 2016-03-15 01:00:02 CETLOG:  server process (PID 22279) exited with exit code 1
>>> 2016-03-15 01:00:02 CETLOG:  terminating any other active server processes
>>> 2016-03-15 01:00:02 CETWARNING:  terminating connection because of crash of another server process
>>> 2016-03-15 01:00:02 CETDETAIL:  The postmaster has commanded this server process to roll back the current
transactionand exit, because another server process exited abnormally and possibly corrupted shared memory. 
>>> 2016-03-15 01:00:02 CETHINT:  In a moment you should be able to reconnect to the database and repeat your command.
>>> 2016-03-15 01:00:02 CETLOG:  all server processes terminated; reinitializing
>>> 2016-03-15 01:00:02 CETLOG:  could not remove shared memory segment "/PostgreSQL.1804289383": No such file or
directory
>>> 2016-03-15 01:00:02 CETLOG:  semctl(22839296, 0, IPC_RMID, ...) failed: Invalid argument
>>> 2016-03-15 01:00:02 CETLOG:  semctl(22872065, 0, IPC_RMID, ...) failed: Invalid argument
>>> 2016-03-15 01:00:02 CETLOG:  semctl(22904834, 0, IPC_RMID, ...) failed: Invalid argument
>>> …
>>>
>>> I am calling pg_dump in my cronjob like this:
>>> su - mmppostgres -c "/Users/…/bin/pg_dump -p 5433 mydatabase_1_0_0 > /my_backup_path/mydatabase_1_0_0.dump"
>>>
>>> After the crash the database runs in recovery mode. A restart of the database brings everything back to normal.
>>> This crash is always reproducible and occurs every night during backup. When calling the same command via the
commandline everything run fine. In the system log I can see: 
>>>
>>
>> Is the command you run via the command line exactly the same, including the su -?
>
> Yes.
>
>> What user are you running the cronjob as?
>
> root
>
>> How do you supply the password for the mmppostgres user?
>
> I configured in pg_hba.conf:
>
> local  all  mmppostgres  trust
>
> All local connections from this user are trusted.

So what happens if you either?:

1) In the root crontab, change the command to:

/Users/…/bin/pg_dump -p 5433 mydatabase_1_0_0 -U mmpostgres > /my_backup_path/mydatabase_1_0_0.dump

2) Run the command in 1) in the mmppostgres crontab

>
> Best regards
>
> Matthias Schmitt
>
> magic moving pixel s.a.
> 23, Avenue Grande-Duchesse Charlotte
> L-3441 Dudelange
> Luxembourg
> Phone: +352 54 75 75
> http://www.mmp.lu
>
>
>
>
>


--
Adrian Klaver
adrian.klaver@aklaver.com