Hi,
I am testing a storage array with iometer in windows Server 2012 Datacenter.
Initiator: Windows iSCSI initiator.
Performed a 2 step test:
1)Removed the network cable from storage node for 10 seconds
2)During this time, I restarted iscsi target services
iSCSI Lun got connected back immediately after that.
But for next 25 seconds, I was getting a target BUSY response from node.
This implies that the LUN was unavailable for just 35 seconds
After that IOMeter started giving errors continuously.
Message seen in Dynamo:
Error performing I/O to #device.
No I/O error was seen at the storage side.
Timeout values for iSCSI initiator as seen in registry are:
DelayBetweenReconnect: 5
EnableNOPOut: 0
ErrorRecoveryLevel: 2
FirstBurstLength: 65536
ImmediateData: 1
InitialR2T: 0
IPSecConfigTimeout: 60
LinkDownTime: 15
MaxBurstLength: 262144
MaxConnectionRetries: 4294967295
MaxPendingRequests: 255
MaxRecvDataSegmentLength: 65536
MaxRequestHoldTime: 60
MaxTransferLength: 262144
NetworkReadyRetryCount: 10
PortalRetryCount: 5
SrbTimeoutDelta: 15
TCPConnectTime: 15
TCPDisconnectTime: 15
WMIRequestTimeout: 30
Disk Timeout Value: 60
If MaxRequestHoldTime is 60, Disk Timeout value is 60 and the iSCSI LUN was reconnected back within 35 seconds, Why is IOMeter giving these errors ?
Is there something I am missing out in my analysis ?
Or, Any other value is used by IOMeter for I/O timeouts.
--All the above analysis was done using wireshark.
Thanks.
IOmeter is not aware of nor does it issues any explicit commands to the underlying iSCSI devices. It only understands local disks. So, whatever errors yo are seeing are caused by underlying software.
Regards,
Ved
Related
Bugs: #82