https://tracker.ceph.com/https://tracker.ceph.com/favicon.ico2012-03-22T10:47:06ZCeph Ceph - Feature #2198: add an option to force a down osd to be marked immediately outhttps://tracker.ceph.com/issues/2198?journal_id=92482012-03-22T10:47:06ZSage Weilsage@newdream.net
<ul></ul><p>What's the motivation for doing that? Is it any better than setting the out interval to be something very short?</p> Ceph - Feature #2198: add an option to force a down osd to be marked immediately outhttps://tracker.ceph.com/issues/2198?journal_id=92492012-03-22T11:00:41ZGreg Farnumgfarnum@redhat.com
<ul></ul><p>It guarantees that you always have the set number of copies on-disk when you get a commit, instead of probably having the correct number of copies.</p> Ceph - Feature #2198: add an option to force a down osd to be marked immediately outhttps://tracker.ceph.com/issues/2198?journal_id=92522012-03-22T11:08:42ZSage Weilsage@newdream.net
<ul></ul><p>Not really, a write will still go to N-1 replicas until the new one is backfilled up through the object's position.</p>
<p>In practice, i think this just means backfill may start and then be canceled/discarded if the osd comes back quickly. And there may be one less osdmap published..</p> Ceph - Feature #2198: add an option to force a down osd to be marked immediately outhttps://tracker.ceph.com/issues/2198?journal_id=92532012-03-22T11:17:38ZGreg Farnumgfarnum@redhat.com
<ul></ul><p>Hmm, yeah, I forgot about that.<br />Somebody was asking about it; I'm not sure if they cared exactly but I'm sure there will be people who want to be able to guarantee a number of on-disk copies.</p> Ceph - Feature #2198: add an option to force a down osd to be marked immediately outhttps://tracker.ceph.com/issues/2198?journal_id=100672012-05-01T12:38:19ZSage Weilsage@newdream.net
<ul><li><strong>Tracker</strong> changed from <i>Bug</i> to <i>Feature</i></li></ul>