Red Hat Bugzilla – Bug 189517
ConnFailTimeout in /etc/iscsi.conf is inaffective
Last modified: 2008-04-07 01:03:32 EDT
Description of problem:
While performing 10 second short cable pulls with ConnFailTimeout=30, EMC
PowerPath reported path failures and tresspassed LUNs. If you
cat /sys/class/scsi_host/host6/connfail_timeout it is set to 5 and should be
30; therefore, iscsi_sfnet sees the path failure and passes it up to EMC
PowerPath before teh specified 30 seconds.
Version-Release number of selected component (if applicable):
Steps to Reproduce:
1. set ConnFailTimeout to 30 in /etc/isci.conf
2. cat /sys/class/scsi_host/host6/connfail_timeout; it still is equal to 5
3. run 10 second cable pulls
iscsi_sfnet reports failed paths to EMC PowerPath and EMC PowerPath marks them
DEAD and tresspasses the LUNs. After the path returns EMC PowerPath marks them
as ALIVE and tresspasses the LUNs back.
iscsi_sfnet should log the session failure and hold off passing the path
failure to EMC PowerPath until the 30 second timer has elapsed.
See attached syslog of failing scenario.
Also, I was unable to provide driver debug information via the initscripts
debug level switches because iscsi_sfnet debug is no longer working. The daemon
and driver start; however, the driver does not fully initialize.
Created attachment 128046 [details]
Syslog snipet of 10 second cable pull failure
Accidental duplicate of 189512.
*** This bug has been marked as a duplicate of 189512 ***