high queued -i/o's in powermt output

make sure to have these flags:

OS2007, SC3, SPC-2

if FAs are already in use, set these via hba_flags command.

yes, initiator needs to log-off and log back-in into FA

Refer to the support matrix and look for director bits that align with your environment as long as it qualifies.. Also, download the host connectivity guide and search for "FA bits". This information and tools are availble via PowerLink. Thanks

Hi Dynamox .. I would be following the below steps

1) Do the zoning ( add new FA's 8a:1 and 9a:1 to the cluster)

2) set the SPC2 bit using symmask hba_flag option

3) Drop the Zoning ( between cluster and 8a:1 and 9a:1)

4) Reconnect again with zoning

5) Map the devices to 8a:1 and 9a:1

6) Mask them

7) rescan on the cluster

Finally the cluster should see 4 paths.

Please correct me if i am missing.

If i follow this process no need to Reboot the Cluster right ? 

Also can someone let know where can I find the docs which has info regarding the bit settings for different OS..

the o/p of the first command is from symstat. What is the command u used for the FA port utilization?

The third o/p , if it is from `powermt watch` display the real time values for Queued IOs.

I used symstat -sid <> -type PORT -dir ALL -port ALL -i 10.

I added two more FA pair to the windows cluster 2008. what should we run on the Host so that it sees 4 paths to each device. Will there be any impact on the production cluster powerpath running at 5.2. Also I was referring to Primus emc202413 which says some possible data corruption when adding a new path. can anyone please clarify. What are the steps i need to follow so that the cluster 2008 sees new paths without any issue( very important).

Thanks,

Scannning/adding more paths are online/transparent to the hosts, in UNIX environment. Not sure about the Windows..

And when you do get your extra paths hooked up, please let us know how it goes.  I know I have money on you seeing a big improvement.

Hi Quincy,

From storage side i added two FA paths for the windows 2008 cluster . Now can you please let me what to do on the windows 2008 cluster side so that it sees the new paths. I know we need to do a rescan and after that how do we bring them under powerpath control. For unix hosts we use powermt config, how abt in windows 2008 ? .And also I was referring to the primus emc202413 in that it was mentioned when we add paths to a windows 2008 cluster which has 5.2 as powerpath data corruption may occur.Unfortunately our cluster is running at 5.2.  So, how to get these additional FA  paths on the server side without running into any issues.. any suggestions would be great..

Device  Cap(MB)  Attr  Dir:P
------  -------  ----  ----
0100     207158   (M)   8A:1, 9A:1, 8C:1, 9C:1
0106    1001261   (M)   8A:1, 9A:1, 8C:1, 9C:1
0123     172631   (M)   8A:1, 9A:1, 8C:1, 9C:1
0128     310736   (M)   8A:1, 9A:1, 8C:1, 9C:1
0131     207158   (M)   8A:1, 9A:1, 8C:1, 9C:1
0137     379789   (M)   8A:1, 9A:1, 8C:1, 9C:1
0142     138105   (M)   8A:1, 9A:1, 8C:1, 9C:1
0146     207158   (M)   8A:1, 9A:1, 8C:1, 9C:1
0D23       5395   (M)   8A:1, 9A:1, 8C:1, 9C:1

if the OS can see them, PowerPath will take control of them automatically (on Windows). Run "powermt display dev=all" ..maybe PowerPath already managing the new paths. If not, are you running active/active cluster nodes ? Can you move resources from one to the other and bounce one node at a time ?

What abt primus emc202413 ?

In that its mentioned about upgrading to powerpath 5.3 for a windows 2008 cluster if its running at 5.2 to see new additinal paths.. can you please clarify.

Sorry, i don't have a test environment to verify that for you. Use the workaround documented in the solution and in the mean time schedule PowerPath upgrade.

The MPIO driver bundled with PowerPath 5.2 has a memory leak issue and you really need to upgrade to PowerPath 5.3.x where the new MPIO driver fixes the memory leak.  This leak can cause performance issues especially in clusters.