On Mon, Oct 5, 2015 at 10:40 PM, Serg M <it.sergm@xxxxxxxxx> wrote:
> What difference between memory statistics of "ceph tell {daemon}.{id} heap
> stats"
Assuming you're using tcmalloc (by default you are) this will get
information straight from the memory allocator about what the actual
daemon memory usage is.
> , ps aux | grep "ceph-"
This tells you what the OS thinks your process' memory usage is...
> and "ceph {daemon} perf dump mds_mem"?
This will give you an estimate the daemon is building from
/proc/self/status but not using for much. In the future it might be
useful, but for now I don't think you can/should do much with it.
> Which is better to use for monitoring a cluster?
> Maybe there is better way to check more precise memory usage of ceph daemons
> with a help of python scripts(i'm using them to create own monitoring
> system)?
I'd use the heap stats (which I think you can also get out of the
admin socket) provided by tcmalloc and monitor the amount of memory
used and the amount of unused memory allocated. If the latter grows
too large you might want to instruct it to release memory back to the
OS (this shouldn't be necessary but some users have had issues with
it).
-Greg
_______________________________________________
ceph-users mailing list
ceph-users@xxxxxxxxxxxxxx
http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com
