On 14/05/14 13:24, Christian Eichelmann wrote: > Hi Ceph User! > > I had a look at the "official" collectd fork for ceph, which is quite > outdated and not compatible with the upstream version. > > Since this was not an option for us, I've worte a Python Plugin for > Collectd, that gets all the precious informations out of the admin > sockets "perf dump" command. It runs on our productive cluster right now > and I'd like to share it with you: > > https://github.com/Crapworks/collectd-ceph > > Any feedback is welcome! Fun, I'd just implemented something very similar! I've just pushed my version upstream to: https://github.com/dwm/collectd-ceph There appear to be some minor differences between our designs: * I don't require that a types DB be kept up to date and consistent; rather, I've reused the generic 'counter' and 'gauge' types. * My version includes some historical processing to allow for the calculation of per-period, rather than global, average values. * Your version is nicer in that it communicates with the admin socket directly; I was lazy and simply invoked the `ceph` command-line tool to do that work for me. It's not currently a significant performance hit, but should be improved. I've been feeding all of my Ceph performance counter data to a Graphite cluster via CollectD, using the CollectD AMQP plugin and a RabbitMQ cluster as an intermediary, and Grafana as a query/graphing tool. Apart from causing some stress on the disk-subsystem attempting to write-out all those metrics, this has been working out quite well... Cheers, David -- David McBride <dwm37 at cam.ac.uk> Unix Specialist, University Information Services
 |