Hi Cephers,
Some of us have been observing this issue for a while, see:
https://github.com/ceph/ceph/pull/18494
http://tracker.ceph.com/issues/22510
And still no perfect solution has been suggested.
Here is another trivial one that's probably is far from being ideal. But
we has to deal with the issue somehow as it might cause real failures
with customers.
The idea is to restrict 'hint' advance within some store space region
('unprotected area'), e.g. 80% of the store.
Once hint value jumps out of that space on hint advance - it's reset to
the zero. Allocations from the 'protected' area (rest 20%) are still
allowed but limited to the case when 'unprotected' area is unable to
serve the request. This is achieved automatically by starting the looku
This way sequential allocations we now benefit from are still available
most of the time. And space for large continuous allocations is much
less utilized and hence is better preserved too.
What do you think?
Thanks,
Igor
--
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
