Autofailover and cache?

[Robotbeat]

Does open-e implement fully synchronous or memory synchronous replication with DRBD? I ask this because I'm trying to figure out if it's worth it to stuff a HA pair with system RAM. In other words, does adding lots of system RAM actually help when you are planning on only using autofailover?

Also, how's NAS (NFS) autofailover coming?

[To-M]

The replication is synchronous but not the cache. Replication of the cache on some technologies is very expensive $15k+++.

NFS or the NAS Auto Failover is planned for this year - most likely later part. This is very complicated to do so we need more time.

WB
Write-back cache. This functionality improves data writing performance. Writing is considered completed as soon as the data is stored in the disk cache. Later the disk cache commits the data to disk.

Note

Please note enabling write-back cache may cause data corruption in case of a system crash (e.g. after power failure). Corruption may also appear during failover processing. We strongly recommend using write-through cache if you are transmitting very important data.

If RO and WB are disabled (RO and WB are unchecked) then Write-through cache is used by default. This means that data is not stored in cache. Instead, all data is considered written after it is committed to disk.
Also, how's NAS (NFS) autofailover coming?

[Robotbeat]

Here's the user guide for drbd:
"Protocol B. Memory synchronous (semi-synchronous) replication protocol. Local write operations on the primary node are considered completed as soon as the local disk write has occurred, and the replication packet has reached the peer node. Normally, no writes are lost in case of forced fail-over. However, in the event of simultaneous power failure on both nodes and concurrent, irreversible destruction of the primary's data store, the most recent writes completed on the primary may be lost."

This would be good for where you have the two boxes on separate UPSes and need high performance (lower latency than "fully synchronous"). It sounds like you wouldn't have a single point of failure even with the memory-synchronous mode.

I understand that open-e's implementation of DRBD is not always completely up to the latest, bleeding-edge drbd release because it needs to be stress-tested and integrated into the GUI. But, it sounds like this is something that should be available sometime from open-e.

[To-M]

Keep in mind that we will also have asynchronous volume replication in next comming release in April.