Symantec Veritas Cluster Server One, Installation Manual

Page: 158 / 210

158 Troubleshooting
Troubleshooting I/O fencing
How I/O fencing works in different situations
Table C-1 describes how I/O fencing works to prevent data corruption in
different situations. The table also includes actions the operator can take in
each situation.
Table C-1 I/O fencing scenarios
Event Node A activity Node B activity Operator action
Both private
networks fail.
Races for majority of
coordinator disks.
If it wins race for coordinator
disks, Node A ejects Node B
from the shared disks and
continues.
Races for majority of
coordinator disks.
If it loses the race for the
coordinator disks, Node B
removes itself from the Policy
Master cluster.
When Node B is ejected
from the VCS One cluster,
repair the private
networks, then try to bring
Node B back.
Both private
networks
function again
after event above.
Continues to work. Crashes. It cannot start the
database because it cannot
write to the data disks.
Restore the private
networks and reboot Node
B.
One private
network fails.
Prints message about an
IOFENCE on the console but
continues.
Prints message about an
IOFENCE on the console but
continues.
Repair the private
network. After network is
repaired, both systems use
it automatically.
Node A hangs. Extremely busy for some reason
or is in the kernel debugger.
When Node A is no longer hung
or in the kernel debugger, any
queued writes to the data disks
fail because Node A is ejected.
When Node A receives message
from GAB about being ejected,
it removes itself from the Policy
Master cluster.
Loses heartbeats with Node A,
and races for a majority of
coordinator disks.
Wins race for coordinator disks
and ejects Node A from shared
data disks.
Verify that the private
networks function and
reboot Node A.