On Tue, Sep 11, 2012 at 6:01 AM, Edward Ned Harvey (lopser) <lop...@nedharvey.com> wrote:
> Ideally, I'd love to see something comparable to zfs mirroring with an iscsi > device. That is - If the iscsi device disappears and reappears, it only > needs to resilver the blocks that changed in the meantime. And the iscsi > device is never inconsistent. (Both characteristics lacking by drbd). DRBD makes great strides to only shuffle changes bits from the bitmap after a split event. I don't think I've ever had to do a full boat re-sync of an existing replica pair. I have done oddball things where I've moved a fs to a standalone DRBD, rebuild the source with DRBD, and then reshuffled everything back. That did take a few hours of pain, but was expected. Generally, DRBD resyncs perform as you indicate you would like them to. "never inconsistent" seems like a slippery slope. If you're using DRBD mode C and a properly setup journaling filesystem, any kind of disruption to the sync should be well fixed at reconnect via bitmap change. In the case where the secondary system gets disconnected and is forced into the primary role, the fs journal should bring up the fs in the same state as it was when the previous primary crashed, and in theory you'll only loose the mid-write transactions on the FS. I've had a number of planned and unplanned DRBD failovers recently and have not found any corruption or data loss. That said, this is still synchronous replication of a live filesystem, which is very different from "zfs send"ing a snapshot as a replica copy. -n -- ------------------------------------------- nathan hruby <nhr...@gmail.com> metaphysically wrinkle-free ------------------------------------------- _______________________________________________ Tech mailing list Tech@lists.lopsa.org https://lists.lopsa.org/cgi-bin/mailman/listinfo/tech This list provided by the League of Professional System Administrators http://lopsa.org/
