We discussed this forever ago during the Giant CDS. I put together a patch set tonight that implements it in a pretty basic form. See https://github.com/ceph/ceph/pull/3429 There's no fancy threading do to this in the background or any of the other more crazy stuff we discussed before. Instead, there's just a simple limit that bounds the amount of time the primary is allowed to spend calculating mappings (default 500ms). It also tries to be reasonably smart about limiting the amount of work, mainly by keeping a mapping of osds to pgs so that we can limit our checks when osds go down or out. My main concern is whether the pg_by_osd map will consume much CPU or memory to maintain. Anybody interested in testing that out? sage -- 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
