Re: osd out states

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



I would definitely support the false/true/true defaults over
true/true/true. Nothing is worse than trying to troubleshoot a
half-dead server in a web cluster and having it put itself back into
the load pool every time it restarts or reloads.

Matthew

On Wed, Jan 4, 2012 at 19:06, Sage Weil <sage@xxxxxxxxxxxx> wrote:
> I pushed branch wip-osd-out-states that adds a few new osd state bits to
> the osdmap:
>
>  * CEPH_OSD_AUTOOUT: osd was marked out automatically by the monitor
> because it was down for too long
>  * CEPH_OSD_NEW: osd is new and has never been marked in
>
> Both are implicitly cleared when the osd_weight becomes non-zero (i.e.
> when they are no longer out).  The AUTOOUT is set by the monitor when it
> automatically marks the osd out.
>
> There are then some new monitor options to go along with
>
>  * mon osd auto mark in (default true): mark osd in when it boots
>  * mon osd auto mark auto out in (default true): mark AUTOOUT osd in when
> it boots
>  * mon osd auto mark new in (default true): mark NEW osd in when it boots
>
> I would imagine an admin might want the first to be false (so that an osd
> they mark out stays out, even if the daemon restarts).  In fact, we should
> make that default to false, I think.  This is true for the teuthology
> thrashing tests, which explicity control the up/down and in/out state of
> OSDs.
>
> For a new cluster, you want new osds to automatically get data.  For an
> existing cluster, you might want that off, so that when you bring a new
> rack of storage online you can explicitly control when you want to start
> using it (i.e., not necessarily when you first boot things up).
>
> The autoout one will likely ways be true, since it is probably safe to
> automatically mark something in that we automatically marked out.
>
> Anyway, comments welcome!
>
> sage
>
> --
> To unsubscribe from this list: send the line "unsubscribe ceph-devel" in
> the body of a message to majordomo@xxxxxxxxxxxxxxx
> More majordomo info at  http://vger.kernel.org/majordomo-info.html
--
To unsubscribe from this list: send the line "unsubscribe ceph-devel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html


[Index of Archives]     [CEPH Users]     [Ceph Large]     [Information on CEPH]     [Linux BTRFS]     [Linux USB Devel]     [Video for Linux]     [Linux Audio Users]     [Yosemite News]     [Linux Kernel]     [Linux SCSI]
  Powered by Linux