Re: Multi-filesystem wthin a cluster

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

 



Hi Marvin,

On Thu, Jan 17, 2019 at 12:36 AM Marvin Zhang <fanzier@xxxxxxxxx> wrote:
> 1. In current design, system cannot create more than 1 fs in a pool.
> In Feature #15066(http://tracker.ceph.com/issues/15066) will handle
> this issue.  Filer::write() will identify the RADOS object by
> poolID+OID. If we want to support multi-fs in a pool, we need to add
> prefix tag on OID, such as fs1:100000000 .PoolID  still comes from
> layout information, which comes from mdsmap. Is it correct?

Yes. See this old ticket: http://tracker.ceph.com/issues/5520

The kernel client would also need updated.

> 2. If the Feature #15066 resolved, can sinlge active MDS support
> multiple FS? In my opinion, an active MDS has a MDSRank object, which
> has a MDLog. We need to change that MDSRank object has multi MDLog.
> One MDLog correspond one fs. It seems no issue tracking this problem.

If I understand your question correctly, we have no current plans to
make a single MDS take on ranks for multiple file systems. Although
attempts have been made in the code -- when easy -- to make this
feasible in the future. This would probably be a significant amount of
work with dubious benefit in the current age of containerized
services.

-- 
Patrick Donnelly



[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