On 05/31/12 11:46 AM, Beth Albertson wrote:
Has anyone ever actually implemented a replicating iSCSI SAN using CentOS? I saw directions here:
http://wiki.centos.org/HowTos/Ha-Drbd
It looks doable, but I am not the strongest linux person and would like to hear if anyone has tried and if there any caveats.
SAN infrastructure should be built to be ROCK solid, as OS's really don't like their block storage devices turning flakey on them. you'd want to use synchronous DRBD replication , and I have no idea how you'd mirror any write-back cache of the two servers in the failover cluster. without some form of writeback caching, ISCSI will be SLOW... ditto, DRBD sync replication is relatively slow (master target has to wait for slave target to commit writes before it can reply to the initiator that the data is committed).
I've had very mixed results from ad-hoc ISCSI tests I've done, and have avoiding deploying any such in production, even for development lab purposes, due to this.
I'd not run /any/ block storage infrastructure on a virtualized platform, that just adds more points of failure, and lowers performance.