Bug #5079
open
panicstr = stmf_svc: I/O deadman hit
0%
Description
(moved here from site triage:)
I experienced following panic on two OI-boxes simultaneously, however these two servers do not share any i/o systems but they do share identical workload, therefore I'm suspecting software issue.
"panic message: stmf_svc: I/O deadman hit on STMF_CMD_LU_OFFLINE after 1000 seconds"
stmfadm was called before problem was exposed. No offline/remove-related procedures though, just adding few luns and views.
Servers are Dell R720 and crashdump is found at http://www.mui.fi/vmdump.0
Other system that had same issues (different datacenter, similar setup) http://www.mui.fi/vmdump.1
As workaround I disabled stmf deadman by: echo stmf_io_deadman_enabled/W0 | mdb -kw
I suspect this could be related in https://illumos.org/issues/3621 ?
Thanks for your thoughts!
Updated by Markus Kovero about 9 years ago
panicstack = stmf:stmf_wait_ilu_tasks_finish+b8 () | stmf:stmf_svc+112 () | genunix:taskq_thread+285 () | unix:thread_start+8 () |
Updated by Markus Kovero about 9 years ago
Fourth system affected, disabling stmf_io_deadman did help, system did not panic at this time, but sbdadm delete-lu gives resource busy.
https://illumos.org/issues/3621 is definitely related.
Updated by Rich Lowe almost 9 years ago
- Project changed from OpenIndiana Distribution to illumos gate
- Category deleted (
OS/Net (Kernel and Userland)) - Target version deleted (
oi_151_stable)
Updated by Markus Kovero almost 9 years ago
Problem seems to be fixed in OI Hipster with illumos-fe2e029.
Updated by Markus Kovero over 8 years ago
Unfortunately, problem still exists in illumos-f8554bb, although it happens much less frequently than before. I'm uploading crashdump currently.
Updated by Markus Kovero over 8 years ago
Latest crashdump can be found from http://www.mui.fi/vmdump.0 original url's no longer have old dumps.