5 milliseconds of latency should not be a problem for the monitorOn Tue, Jul 23, 2013 at 8:54 AM, Matthew Walster <matthew@xxxxxxxxxxx> wrote:
> I've got a relatively small Ceph cluster I'm playing with at the moment, and
> against advice, I'm running the MONs on the OSDs.
>
> Ideally, I'd like to have half the OSDs in a different facility and
> therefore have one MON in each facility.
>
> Which gives the problem of where to put that pesky third MON...
>
> Is there any likely performance hit if I was to put the third MON in a
> totally different datacenter, potentially in a different city? I know OSDs
> need to be able to complete a transaction across all nodes in a placement
> group before finishing a write, but what about MONs?
>
> As I understand it, the MONs just describe the topology, so surely having
> them 5ms away shouldn't be that big of a deal?
>
> Just wondered what people's thoughts were.
workload. You'll want to be sure and direct all the monitor clients to
the monitor in their local DC though[1] or you'll get a lot more
bandwidth going across DCs than is necessary.
That's fantastic, thanks. I'm assuming that 5ms is probably too much for the OSDs -- do we have any idea/data as to the effect of latency on OSDs if they were split over a similar distance? Or even a spread - 0.5ms, 1ms, 2ms etc. Obviously this is a bit theoretical as you'd not want to have data that could be local pulled from a far away data center over your expensive links when in-datacenter would clearly be the better choice.
Matthew Walster
_______________________________________________ ceph-users mailing list ceph-users@xxxxxxxxxxxxxx http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com
