I know this an old comment, but we just ran into something similar. We
had entries for pmlogger just in plain old /etc/crontab, also with -k.
We have been running pcp for many years so I'm not sure what the origin
of this file is, might be our own doing. On an rpm upgrade, the new
/etc/cron.d entries are installed. With both set to run at 12:10, from
what I can tell, its random which version wins. I'm note sure if
anything can be done about this, and we should have been more careful on
upgrade, but just another data point for a possible issue on these files.
Martins
On 11/20/13 5:14 PM, Nathan Scott wrote:
----- Original Message -----
We install crontab entries, e.g.
/etc/cron.d/pcp-pmie
/etc/cron.d/pcp-pmlogger
but we used to install them in different places and/or with different names.
There is some effort to handle migration during upgrades with a mixture
of rules and "pre" scripts.
I have just observed a spectacular failure of this.
In upgrading a Centos system from PCP 3.8.0 to PCP 3.8.9,
/etc/cron.d/pcp-pmlogger existed before the upgrade (and had been
modified), but after the upgrade a new /etc/cron.d/pcp-pmlogger was
installed without any warning, and no .rpmnew, ... assistance. So the
old /etc/cron.d/pcp-pmlogger was lost.
These are marked as noreplace configuration files...
# rpm --query --configfiles pcp | grep cron
/etc/cron.d/pcp-pmie
/etc/cron.d/pcp-pmlogger
so, not clear how this happened. Were these files that had been created
(outside of PCP packaging) before we begun installing crontabs? i.e. is
this a coincidental naming collision, and rpm didn't know about 'em?
I suppose things are tricky in the cron.d directories because we cannot
leave around .rpmsave files and so on - both variants would still be found
by cron. Perhaps (not sure) rpm has different logic for these directories
to ensure duplicates don't happen.
cheers.
--
Nathan
_______________________________________________
pcp mailing list
pcp@xxxxxxxxxxx
http://oss.sgi.com/mailman/listinfo/pcp
|