Fp_offline_ticker And Fcp_offline_delay Not Effective @ Solaris 11.2.3.4.1 (Doc ID 2002384.1)

Last updated on AUGUST 25, 2015

Applies to:

Solaris SPARC Operating System - Version 11.2 and later
Information in this document applies to any platform.

Symptoms

On a host running Solaris 11.2.3.4.1, with fabric connections,
and mpxio installed, there are two host bus adapter connections
to the switch connected to the storage.

When testing (temporary) removal of both host bus adapter
connections to the switch between the host and the storage, while
application IO is active (including writes), errors are immediately
generated.

......
Jan 20 11:12:31 qlc: [ID 628150 kern.info] NOTICE: Qlogic qlc(1,0,1): Link OFFLINE
Jan 20 11:12:50 qlc: [ID 628150 kern.info] NOTICE: Qlogic qlc(0,0,0): Link OFFLINE
Jan 20 11:12:51 ufs: [ID 702911 kern.warning] WARNING: Error
writing master during ufs log roll
Jan 20 11:12:51 ufs: [ID 127457 kern.warning] WARNING: ufs log for /mnt changed state to Error
Jan 20 11:12:51 ufs: [ID 616219 kern.warning] WARNING: Please umount(1M) /mnt and run fsck(1M)

Jan 20 11:13:34 qlc: [ID 628150 kern.info] NOTICE: Qlogic qlc(1,0,1): Link ONLINE
Jan 20 11:13:36 qlc: [ID 628150 kern.info] NOTICE: Qlogic qlc(0,0,0): Link ONLINE
......

The delays set by fp_offline_ticker (default 90 seconds) and
fcp_offline_delay (default 20 seconds) are not honoured. Both
of these fcp and fp delays are ignored, and we do not get any delay
before errors are returned to the application layer.

The objective was to remove, and replace, both connections to the switch
after a delay of less than fp_offline_ticker + fcp_offline_delay seconds
(default 110 seconds in total), to test resilence, and make sure that

temporary, but brief, connection interruptions could be survived.

Cause

Sign In with your My Oracle Support account

Don't have a My Oracle Support account? Click to get started

My Oracle Support provides customers with access to over a
Million Knowledge Articles and hundreds of Community platforms