No failover when volume hardddisk fails?

[Christian Pelg]

Hi everyone,

I have done some tests on open-e dss v6 trial. Besides other things I wanted to see what the reaction is when I unplug the harddisk during use that contains the iSCSI volume. The result is that there is no failover to the secondary node happening on my system. Strange is that the system shows an error message (seems to realize the problem), but doesn't do the failover to the secondary node. I am using a single harddisk (no RAID) for my tests and I understand that later with RAID the problem wouldn't be that critical, but then again even a complete RAID array can fail and then there would be the same problem again. I expected another reaction than this on a system that promises high availability. When I unplug the harddisk that contains the open-e system then there is no reaction either. Of course in that case the systems keeps running and stays fully functional in that case as if it was running from CD. Open-e DDS really just seems to handle problems as expected that involve NICs or the loss of power of the whole system. My setup and configuration are following the "Step-by-Step Guide to Synchronous Volume Replication (Block Based) with Failover over a LAN supported by Open-E DSS".

Thanks in advance for your opinion.

Christian

[transparent]

That seems normal - our failover only occurs if the primary node stops responding to pings (ie complete failure of the entire box). I do not believe it is monitoring the actual iSCSI service, or if data is available. It would be great if the failover could monitor a hidden file or volume on the primary, and if that file or volume became inaccessible, it would fail over.

It a very low level monitoring, does not actually reflect the state of the service.

[Christian Pelg]

Thank you very much for answering. At least I know now that this seems to be normal system behavior and it's not because of my setup, configuration or whatever.

I would still like to hear more experiences regarding the detection of harddisk failure.

[enealDC]

Best thing you can do is probably respond to an snmp trap or alert and power off outlet at the PDU...

[Christian Pelg]

I've contacted open-e concerning this and they agree that harddisks are not monitored yet. They said though that they are working on this and harddisk monitoring will probably be included in the new open-e release coming in September.