On Fri, Jun 22, 2018 at 5:44 AM, Lars Marowsky-Bree <lmb@xxxxxxxx> wrote:
> On 2018-06-20T09:25:30, Alfredo Deza <adeza@xxxxxxxxxx> wrote:
>
>> Kind of. Bluestore OSDs do not need to have a block.db and block.wal
>> defined. In this case the whole device would be 1 LV
>> which will allow bluestore to use it without external db or wal.
>>
>> That is, ceph-volume will *not* create 3 LVs from {device} for 1 OSD,
>> it will create 1 LV from {device} and place a single bluestore OSD in
>> there.
>
> Would it make sense if it did? Then one could "easily" use standard
> pvmove to shift them to a different backing device should the need
> arise?
It depends, if block.db starts spilling over to the data LV when it
runs out of space it gets quite complicated to attempt to move that
out and continue with the performance benefits. IMO it is far easier
to just do the OSD again.
To prevent this from happening, the block.db LV has to be sized
accordingly, but this is hard to do without a good understanding of
cluster workloads (e.g. RBD or RGW?)
In the current approach we are just carving out block.db LVs as big as
possible from solid devices, but allowing further customization from
higher level tooling (e.g. ansible)
>
>
>
> Regards,
> Lars
>
> --
> SUSE Linux GmbH, GF: Felix Imendörffer, Jane Smithard, Graham Norton, HRB 21284 (AG Nürnberg)
> "Architects should open possibilities and not determine everything." (Ueli Zbinden)
> --
> 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
