-
Verify that all the expected mpxio paths are online using luxadm
display <raw path to lun slice 2>, you can determine the
paths using format.
luxadm display /dev/rdsk/c8t600C0FF00000000000136C6215C0B500d0s2
^ use the raw path!
-
In a window start a #tail -f /var/adm/messages
-
In another window start an #iostat -xnMcz 5
-
In another window start an sccli in band session with the
attached
array.
-
In another window generate some I/O load to all of the luns
presented by the array using the raw mpxio paths. Use 1 dd read
instance per Lun on the slice 2 path. Wait for the iostat output to
stabilize (1 minute).
# dd if=/dev/rdsk/c8t600C0FF00000000000136C6215C0B500d0s2 of=/dev/null
bs=32k &
-
From the sccli prompt fail the primary controller. Make a note
of
which controller has the blinking green LED as it is the present
primary
controller.
sccli> fail primary
The activity may slow down or even seem to stop in iostat, but
it should come back in no more than 2 minutes. Messages may
appear in the messages window, but keep an eye on iostat. The
most important thing is that the I/O continues. Give the system 2
min to stabilize.
#luxadm display /dev/rdsk/c8t600C0FF00000000000136C6215C0B500d0s2
This should show all paths online.
sccli>show inq
(this ensures that the secondary controller can service
sccli)
At this point you have forced a take-over by the default
secondary. You could establish a telnet session to ensure
this controller's network port is working. (They should both
be plugged into the network).
-
This step is a cable pull, only one path will be taken OFFLINE
in this step.
Pull the upper channel 0
cable.
Watch /var/adm/messages for the offline or degraded path
messages.
#luxadm display
/dev/rdsk/c8t600C0FF00000000000136C6215C0B500d0s2
The number of paths should be reduced by half. Logically,
if your number of paths have been reduced to one and iostat
continues to show activity, you really do not need to look at the
paths to know that the one is working.
Again, you should see some messages in /var/adm/messages. Iostat
may have another slowdown or percieved stop. Iostat should
stablize within 2 minutes and bandwidth may be affected by as much
as half. Connectivity is the key. Now plug upper channel 0
back in to ONLINE the path.
Wait 2 minutes for the system to stablize.
Look for the path ONLINE messages in /var/adm/messages
#luxadm display /dev/rdsk/c8t600C0FF00000000000136C6215C0B500d0s2
The number of paths should be back to normal. Iostat
bandwith, if
it was affected, should be back to normal.
-
Repeat step 8 for lower channel 5.
Pull the lower channel 5
cable.
#luxadm display /dev/rdsk/c8t600C0FF00000000000136C6215C0B500d0s2
The number of paths should be reduced by half.
Again, you should see some messages in /var/adm/messages.
Iostat may have another slowdown or percieved stop and should
stablize within 2 minutes. Bandwidth may be affected by as
much as half. Connectivity is the key. Now plug upper channel
5 back in to ONLINE the path.
Wait 2 minutes for the system to stablize.
#luxadm display /dev/rdsk/c8t600C0FF00000000000136C6215C0B500d0s2
The number of paths should be back to normal. Iostat
bandwith, if it was affected, should be back to normal.
-
Deassert (restore) the failed controller from step 7
sccli> unfail primary
wait 5 minutes for the system to stablize.
You may see some activity in /var/adm/messages. Iostat
should be mostly unaffected. If you are watching the array LED's,
the unfailed controller will be yellow during post. Then both
will go yellow for a second and finally the unfailed one will go
solid green showing it is now the secondary.
-
The original secondary controller is now primary and this
procedure has verified it's base functionality. We will now fail
the present primary; making the original primary, "primary" again.
Then we will accomplish all the same tests on the original
primary. To accomplish this, repeat steps 6 through 9
verbatim. When you are done, after you do step 9 again the
se3510 will be in it's original state. Once you are done running
steps 6-9 to test the paths on the primary controller, you are
done.
You can now
#pkill dd
Quit out of iostat and tail, and power cycle the se3510.
