Gluster users - Jul 2016 - self healing with sharding

2016-07-08 11:41 GMT+02:00 Kevin Lemonnier <lemonnierk at
ulrar.net>:
> That's not really what happens, the files won't be set read only as
> gluster has no idea what files are in the VM disk.
> The shards being healed will be locked, not set read only, that means
> the process trying to access a locked shard will just hang, waiting for
> the lock.
> With smaller shards that should be mostly transparent though, since healing
> 64 Mb is very fast.
So, the best way is to figure out the best ratio between shard size
and performance.
Keeping the shard size as small as possible but at the same time
trying to reach the best performance.

If performance would be the same with 10MB shard or 100MB shard, the
10MB is better because allows a faster healing.

On 8/07/2016 8:40 PM, Gandalf Corvotempesta wrote:
> If performance would be the same with 10MB shard or 100MB shard, the
> 10MB is better because allows a faster healing.
Depends on what performance measures you use. In my ad hoc testing, my 
impressions where that large seq read/write slowed down as shard size 
shrunk, but oddly, random i/o actually increased.


For me, the happy medium is 64MB. But it will depend on your workload.

-- 
Lindsay Mathieson