Bug #242
TCP/IP related panic in NCP 3.0
| Status: | New | Start: | 08/27/2010 | |
|---|---|---|---|---|
| Priority: | High | Due date: | ||
| Assigned to: | - | % Done: | 0% |
|
| Category: | - | Spent time: | - | |
| Target version: | - |
Description
I have an X4240 with 16GB of memory running NCP 3.0 with six production zones. It has been crashing randomly severaly times a day over the past week after upgrading to NCP 3.0, with the only indication of a problem being the following panic message:
panic[cpu0]/thread=ffffff0513d27e60: decr_upcount-off the end
ffffff001fabfcf0 genunix:upcount_dec+81 ()
ffffff001fabfd20 genunix:freeproc+71 ()
ffffff001fabfdc0 genunix:waitid+2e0 ()
ffffff001fabfec0 genunix:waitsys32+30 ()
ffffff001fabff10 unix:brand_sys_syscall32+192 ()
This appears related to 6923355.
This machine previously ran NCP 2 and was completely stable. It has several physical NICs, including a Sun multi-threaded 10GbE using nxge and the built-in 4x1GbE using nge. All the zones are currently using the nxge NIC but I will be testing them on nge soon.
If it is the referenced bug then it is unlikely NIC/driver specific.
-phillip
History
Updated by Phillip Steinbachs about 1 year ago
Just for tracking purposes, it looks like this has been fixed in the Illumos repo with revision 11680.
I can reproduce this bug consistently on NCP 3.0.1 by restarting a zone running apache 2.x fronting Tomcat via mod_proxy and have done so on several machines with different hardware.
Updated by Phillip Steinbachs about 1 year ago
Clarification on the above.. I meant to say restarting apache inside of a zone.
Updated by Albert Lee 9 months ago
This is fixed in ncp3-gate: changeset: 274:8e194ec5a7bc parent: 271:c5691a3552e0 user: Albert Lee trisk@nexenta.com date: Thu Feb 24 11:42:58 2011 -0500 summary: re #4054 rb491 Backport important ip stack fixes
Until NCP is at 3.1, you can pull updated packages from http://apt.nexentastor.org/3.1/