From Bugzilla Helper: User-Agent: Mozilla/4.0 (compatible; MSIE 6.0; Windows NT 5.0; EMC IS 55; .NET CLR 1.0.3705; .NET CLR 1.1.4322) Description of problem: RHEL 3.0 U4 host hangs when running port disables on Cisco switch ports for CLARiiON SPs. The host is Dell PE4600 using the on-board e1000 NIC running RHEL 3.0 U4 v2.4.21-27.ELsmp kernel. The driver being used is the integrated 3.6.1 iscsi driver attached to a CLARiiON iSCSI array via a Cisco switch. After a variable amount of time (anywhere from 30 minutes to a day and a half), the host would hang while the port disable script was running ( 5 minute disable with a 10 minute wait). The host was running Iozone (filesystem I/O). Nothing of substance was being reported in/var/log/messages and the host was simply hung so a power cycle was required. Enabled the NMI Watchdog to initiate a panic. When the system hung, it produced a similar panic with PowerPath v4.3.1 (release), v4.3.2 (beta), and without Powerpath at all. Version-Release number of selected component (if applicable): iscsi_sfnet v3.6.1 How reproducible: Always Steps to Reproduce: 1. Attach your RHEL 3.0 U4 host via iSCSI to a CLARiiON array. 2. Run I/O. 3. Disable the switch ports. Actual Results: The host will hang. Expected Results: The host should not hang and should continue running. Additional info:
Created attachment 111180 [details] serial console output from a host with no PowerPath at all Attached is the serial console output from the host running through the port disables without PowerPath.
Created attachment 111181 [details] panic report from a host with PowerPath v4.3.2 Attached is the panic report from the host running through the port disables with PowerPath v4.3.2.
Created attachment 111182 [details] iscsi.conf that was being used In case it's needed, I've attached a copy of the iscsi.conf file that was being used on this host. It was the same for both PowerPath and non-PowerPath testing. Thanks.
Created attachment 111203 [details] panic report from host with PowerPath v4.3.2 Attached is the clean panic report from the host running through the port disables with PowerPath v4.3.2.
In the opennnig comment you said that the on-board e1000 NIC is being used. The crash output indicates that tg3 and e100 are loaded. No e1000. Assuming that this failure was with the tg3, would you be able to re-test with an e1000 NIC, so we can see if the problem is specific to the tg3? Also, would you be able to re-test on tg3 without iSCSI? Just run a network load over the tg3. Thanks. Tom
Per Wayne, this problem is no longer occurring since moving to RHEL 3.0 U5 and using a different iscsi.conf file. Waiting for an update on his testing before closing the bugzilla.
Closing the Bugzilla as the problem hasn't been replicated with RHEL 3.0 U5.