I've got a two node OES2 Linux cluster running iSCSI with a Netware 6.5 SP7 server as the target. The linux servers can discover and login to the target fine however when data is read or written to the target "Bad ExpStatSN Got xxxx expected xxxx" appear on the Netware server logger screen. Also "iscsi: GOT_CHECK_CONDITION but invalid data buffer size of 0" is written to /var/log/messages on the OES servers. The OES2 readme document states:

2.9 iSCSI Initiator for Linux Hangs When Writing Files to a NetWare iSCSI Target

If the NOOP settings in the Linux iSCSI initiator are set, the Netware iSCSI target reports back a SCSI CHECK CONDITION error when it receives the NOOP request during an active iSCSI read or write request sequence. In response, the Linux initiator calls the BUG_ON condition, resulting in a hang. To avoid or resolve this problem, turn off the NOOP settings in the /etc/iscsid.conf file by setting the values to 0. For example:

noop_out_timeout = 0

noop_interval = 0

Afterwards, do a rediscovery in order for the parameters to take effect.

I have done the above and yet the problem persists. I can confirm that writing or reading a large amount of data can cause the kernel to panic. When running NCS you end up with a situation where each cluster node sends the other a poison pill. The problem also occurs when using a NW6.5 SP6 target.

Has anyone experienced this and managed to resolve the issue?