Ahoj Honzo, On 11/08/2014 02:02 AM, Jan Kundrát wrote: > Hi Jiri, > I have to admit that we are doing this on just a couple of DBs, but a > configuration management system sounds like the obvious place of > implementing such a feature. As a bonus, it's usually easier to > maintain and hack together than a C plugin. Any particular reason why > you wanted this to be in collectd itself? you are right and it is a valid alternative for this as well. The downside is that it forces you to keep the list of databases for each instance of postgresql even in cases you don't really want to do that. For example, where DB administration is job of somebody else, yet you want to have the metrics in case of performance problems. The list will inevitably get out of sync soon or later which we want to avoid. We thought about it in munin's way where auto discovering of various resources is very convenient feature.
Anyway, if nobody in our team feels like hacking in C with a forecast of having to maintain the patch in future (i.e. it won't be accepted), we will probably go with "keep the list of monitored DBs in Puppet" way. Cheers Jirka H. _______________________________________________ collectd mailing list [email protected] http://mailman.verplant.org/listinfo/collectd
