[OpenIndiana-discuss] Disk failure detection in OI

[Miloš Dukić]

Hello list,

I have a setup that includes two bare metal server running oi_151a8
and two JBODs. Servers are connected with JBODs so each can see disks
from both JBODs. Total number of disk is 20 (12x Seagate Contelation
ST2000NM0023, 2x Seagate ST200FM0002, 2x Toshiba PX02SMF020). Both
nodes have LSI 9200-8e controller.

Test zpool is created, and contains 3 striped mirrored vdevs. Each
vdev contains one disk from (let's call it) jbod01 and one disk from
jbod02.
For testing purposes, after powering off jbod01 whole system hangs.
Issuing format/cfgadm (other disk related commands) hangs on both
nodes, after few minutes system goes into panic and reboots.
After reboot, system sees only 10 disks and zpool import can import
degradated pool.

My question is:
Is there a way to configure IO so it could after disk failure
unconfigure drives, and keep zpool up and running in degradated state
without rebooting?

Things I tried:
Adding ddi-failfast-supported:true in /kernel/drv/sd.conf
Adding set sd:sd_io_time=5 in /etc/system
Setting un_retry_count=0x3

Thanks in advance,