Error Message

GAB INFO V-15-1-20036 Port h gen c7da09 membership 01
GAB ERROR V-15-1-20050 Protocol error: recv dlv_seq 78868
Kernel panic - not syncing:
assert failed /vcsbuild/builds/vcs50mp3rp2linux/2009-02-10_12.55.00/buildarea/linux/rhel50/x86_64/comms/gab/uts/io/gab/gabrcv.c ln 872: 0 

Stack:

PID: 13350 TASK: ffff810649c4f040 CPU: 14 COMMAND: "lltdlv"
#0 [ffff81065280db98] crash_kexec at ffffffff800b09a0
#1 [ffff81065280dc58] panic at ffffffff8009398d
#2 [ffff81065280dd48] gab_receive at ffffffff88caafc0 [gab]
#3 [ffff81065280de38] gab_receive_port_que at ffffffff88cad6b6 [gab]
#4 [ffff81065280de78] gab_receive_que at ffffffff88cad9c9 [gab]
#5 [ffff81065280de98] gab_lrecv at ffffffff88ca7166 [gab]
#6 [ffff81065280dec8] llt_lrsrv_port at ffffffff88c70515 [llt]
#7 [ffff81065280df18] cleanup_module at ffffffff88c7adc2 [llt]
#8 [ffff81065280df28] llt_deliver at ffffffff88c713a3 [llt]
#9 [ffff81065280df48] kernel_thread at ffffffff8005dfb1 
 

System panic due to a corrupt LLT packet. Corruption caused by faulty hardware.

Enable llt checksums. With llt checksums enabled, llt will attempt to detect and handle situations where a packet is corrupted. The packet will be dropped causing the packet to be resent.

Please note that checksums have no direct bearing on whether a link is "up", unless the heartbeat packets themselves are corrupted at a rate
large enough to prevent even one packet from going through correctly in 16 sec (default LLT peerinactive timeout). If one link goes down, and only one more is left,
GAB will declare jeopardy independent of LLT checksums.

 How to enable checksum:
 ======================
1) Enable checksum at run time on all the nodes using below command
 # lltconfig -K 10

2) Execute below command to confirm that checksum is enabled.
 # lltstat -c
 --
 checksum level: 10
 --
 3) Add the below line at the end of the /etc/llttab to make this checksum change persistent across the reboots.
 # set-checksum 10

 4) Monitor the output of the below commands periodically on all the nodes and see if there is non-zero value for the "badcksum" field.
 # lltstat
 # lltstat -l

 5) If the outputs above indicate bad checksums, collect the outputs from the below commands for analysis by Veritas Support.
 # lltstat -c > /tmp/lltstat_c_nodeXX.out (here, XX is the node number.)
 # lltstat > /tmp/lltstat_nodeXX.out
 # lltstat -l > /tmp/lltstat_l_nodeXX.out

Note that llt checksums are enabled by default beginning with the 5.1SP1 release

Applies To

Redhat Linux 2.6.18-308.13.1.el5.

SF-HA 5.0MP3. Any VCS release prior to 5.1SP1 may be vulnerable. If LLT checksums is disabled, this can happen on any release.