<div dir="ltr"><div><div><div><div><div><div><div><div><div><div><div><div><div>1) I/O error -> Automatic unplug<br></div>We all agreed that in this case, sheepdog has to trigger a cluster reweight.<br><br></div>2) Manual unplug<br>
</div>As of now, this trigger a recovery but not a reweight.<br></div>I think, in most cases, it's not even necessary to start a recover.<br>e.g.<br> I'm unplugging the disk because I'm going to substitute it "now".<br>
</div></div></div> dog node md unplug /mnt/sheep/2 --hold<br></div> <substitude the disk><br></div> dog node md plug /mnt/sheep/2<br></div> Now the recovery can start and restore the missing objects on /mnt/sheep/2.<br>
</div><div> In case the disk size is different from the original, I'll trigger a reweight.<br><br></div>If option '--hold' is not passed, sheepdog uses the default behaviour but it has absolutely to avoid a disk getting full (as we spoke earlier).<br>
</div><br></div><div>Honestly, I find it difficult to think of a case in which I want to unplug a disk without replacing it asap and without changing the weight of the node.<br><br></div><div><br></div><div><div><div><div>
<div><br><br></div><div><div><div><div><div><div><div><div><div><div><div><div><div><div><br></div></div></div></div></div></div></div></div></div></div></div></div></div></div></div></div></div></div></div>