Cron jobs not finishing database dumps (dbdumps)

[chris stuart_1]

I have 3 databases on this server that run dumps automatically from my crontabs. The first one (small database) finishes then starts the 1st half of the second database (huge) and that finshes but the 2nd half of the database never finishes and the 3rd database (small) never runs dumps automatically at all. i think its a time conflict with another job running. thats the only thing i can think of for the dbdumps terminating before they finish. im using HP-UX 10.20 by the way and i have sybase loaded on it too.

[Pete Randall]

Have you checked /var/adm/cron/log to see if the jobs are starting as scheduled? Do they start and not finish or what?


Pete

[chris stuart_1]

I just checked it and we reboot the server every morning at 4am so thats the only thing it had in there was today and we have the cronjobs scheduled to run the dbdumps at night.

[Pete Randall]

By any chance did the reboot occur before the jobs had a chance to finish?

Can you check the cron log tomorrow before rebooting?


Pete

[spex · Accepted solution]

Chris,

Check /var/adm/cron/OLDlog as well as any sybase logs.

Also consider daisy-chaining the backup scripts together so that one kicks off the next before finishing.

PCS

[Pete Randall]

Duh! Of course, check OLDlog!


Pete

[chris stuart_1]

I just got done checking the OLDlog file. i didnt see anything unusual in there. and the reboot goes 5 hours after the first dump runs. it only takes 1.5 - 2 hours to run all the dbdumps manually because we run them manually on the weekends.

[Pete Randall]

Each entry in the log has 3 lines to it:

> CMD: /home/informix/easycopy/easycopy.sh
> informix 24020 c Tue Jul 11 13:45:00 EDT 2006
> CMD: /apps/hols/bin/wb/checkweb.ksh
> informix 24021 c Tue Jul 11 13:45:00 EDT 2006
< informix 24021 c Tue Jul 11 13:45:00 EDT 2006
< root 24019 c Tue Jul 11 13:45:00 EDT 2006
< informix 24020 c Tue Jul 11 13:45:06 EDT
2006

The first line shows the command, the second the user, PID, date and time info preceded by a ">" symbol, the third is the user PID, end date and time, preceded by the "<" symbol. Did you see all four finish, i.e. did you see a "<" line for each?


Pete

[spex]

Chris,

Make sure OLDlog has two entries per script: one for the start and one for the end. Here's how to read it:

> CMD: /usr/sbin/ntpdate a32dom1 > /dev/null # Sync system clock to a32dom1
> root 10769 c Sun Jul 9 06:05:00 EDT 2006
< root 10769 c Sun Jul 9 06:05:02 EDT 2006

The first line lists the crontab entry. The second line indicates that this job was run as process 10769, which started (note the ">" symbol) at 6:05:00 am. The third line indicates that the same job finished (note the "<" symbol) two seconds later. Note that there may be entries for other jobs between the start and end entries.

Confirm you have a pair (">" and "<") of entries for each script.

PCS