On Wed, Aug 23, 2017 at 2:28 PM, Christian Balzer <chibi@xxxxxxx> wrote:
On Wed, 23 Aug 2017 13:38:25 +0800 Nick Tan wrote:
> Thanks for the advice Christian. I think I'm leaning more towards the
> 'traditional' storage server with 12 disks - as you say they give a lot
> more flexibility with the performance tuning/network options etc.
>
> The cache pool is an interesting idea but as you say it can get quite
> expensive for the capacities we're looking at. I'm interested in how
> bluestore performs without a flash/SSD WAL/DB. In my small scale testing
> it seems much better than filestore so I was planning on building something
> without any flash/SSD. There's always the option of adding it later if
> required.
>
Given the lack (for large writes) of double writes with Bluestore that's
to be expected.
Since you're looking mostly at largish, sequential writes and reads, a
pure HDDs cluster may be feasible.
I have a final question. Assume I'll have 96 OSD's of 10TB each and I use erasure coding with k=5 and m=3 with the failure domain of host.
It's my understanding that with erasure coding and host failure domains, the minimum number of hosts required is equal to k+m. Is this correct?
So, if I use 8 hosts with 12 OSDs each this means I can lose any 3 OSD's but can I lose a host, assuming there's enough free space in the pool to accomodate the missing 12 OSD's? Or will the cluster block at this point because there's only 7 hosts which is less than k+m?
If I use 48 hosts with 2 OSDs each then I can lose up to any 3 OSD's and the cluster can recover. And I could lose 1 host at a time assuming the recovery completes before the next host is lost (until there's 7 hosts left or there's not enough free space in the pool). Is this a correct assessment?
Thanks,
Nick
_______________________________________________ ceph-users mailing list ceph-users@xxxxxxxxxxxxxx http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com
