DMX Out-of-Family Microcode Upgrade Timeout Issues

It's in primus solution.. emc58837

Hello, I have been looking around EMC Powerling and the EMC Support Matrix and the Support Matrix and EMC HP-UX Host Connectivity Guide both state the minimum  recommended PVtimeout value is 90 seconds for online Symmetrix operations. With that said the average time for a Symmetrix In-family code upgrade IML where it loads code to all directors at one time and pauses I/O is about 7 to 10 seconds. With a Out of family Code Upgrade the Symmetrix IML time takes a little longer and some of the factors are the size of the Symmetrix, the amount of physical cache.

The following comes from primus solution emc223786 which explains what happens during a Symmetrix NDU:

It is critical that all of the EMC Engineering-recommended operating system timeout values, the HBA driver versions, the HBA driver settings, and switch firmware revision requirements are met. Refer to Powerlink and the E-Lab Navigator tool or check the latest Host Connectivity Guide for these values. Otherwise, contact the EMC Solution Support Center or your local EMC Service Representative for assistance.

• These EMC-recommended driver and firmware versions and the EMC-recommended host and driver settings, reflect the results of extensive qualification testing by EMC Engineering.
   
  
• To confirm that your storage environment is fully qualified for the NDU process you must run the EMCGrab utility on all hosts attached to the Symmetrix. You then need to verify the EMCGrab output using the EMC Host Environment Analysis Tool (HEAT). Refer to solution emc203900. Any warnings in the HEAT output can then be investigated and resolved with assistance from your EMC Service Representative.
  

Gary Bastarache

You cannot set a generic value for Disk timeout. It depends on what you wan't to achieve.

If you wan't fast failover of paths then keep the value low. Fast failover means when you wan't your app to get the response from another path to the LUN ASAP if the actively used path fails. This can cause issues while Upgrades on Switches and Storage Arrays. The app will unnecessarly show errors.

Conversly, keeping high value, lets say >180Sec might stall the application if you encounter a link loss or performence issue.

Whenever changing this value, you should consult with everybody(App vendor DBA/Sys/Storage Vendor).

PV time out in HP-UX LVM(pvchange -t x) level is different than HBA time out in Linux. I have not seen an equivalent pvchange on Linux LVM2. But you have some control over time outs on 2.6 kernel based systems which depend on /sys/devices/pci0000:00/0000:00:03.0/0000:04:00.0/host0/target0:0:0/0:0:0:0/timeout which already simialr to HBA settings you mentond