Mode 31 Dump FAQ

Why does Hitachi need a Mode31 dump?

How can I ensure the collected data is useful?

Environment

• All Hitachi Storage Systems

Answer

You may need to login to the support pages for links below to operate correctly.

• Firstly the most important thing to understand is that a Mode31 dump data is only useful if collected DURING a problem or impact or at very least soon afterwards. 
• Collecting before or after an event will not provide the trace data required so accurate timing and triggering of the Mode31 dump is critical.
• Mode31 dump traces are captured by any of these methods:
  1. It is recommend to use a "Hitachi Remote Operations" (HRO) web console to automatically enable Mode31, save a "Detailed Dump" and disable Mode31.  The dump should also be automatically uploaded to our TUF servers.
     • You may be locally using either "SVP Hitachi Remote Ops Agent" or  standalone "Hitachi Remote Ops Monitor Agent", see instructions for each below:
       • How to Upload a Dump from the SVP to GSC via Remote Ops
       • How to Collect an Auto Dump Using Remote Ops Monitor Agent
     • Note that due to polling intervals we only recommend using the local HRO web console.  Although Hitachi staff can trigger a remote dump request there will be significant delay between the request and the Mode31 dump starting.
  2. Manually the (System Option Mode) SOM31 can be enabled to constantly trace so that later when a "Detailed Dump" is collected the trace buffers are frozen, dumped and resumed.
     • A Customer Engineer (CE) is usually required to manually enable SOM31. See: Enterprise Data Collection

• Note if the system is using replication it is likely that Mode31 dumps will be required from all other systems around the same time also during a problem or impact.
• It is usually reccommended to collect performance export data after the Mode31 dump has completed. 
  • Detailed export records usually cover 24 hours so running the exports immediately after Mode31 dump will usually include the period of concern. See: Performance Data Collection - How to Run the Export Tool

Additional Notes

• Enabling and disabling the mode 31 does NOT collect any data.  It starts the process of placing IO detailed log records into round-robin trace buffers of each MP.
• Just after starting a detailed dump (within the first few minutes) the system will freeze each MP trace buffer and store the detailed logs into the hdcp.tgz dump file. 
• The mode 31 trace log is a very small microscopic point in time typically capturing IO traces for less than 1 second.  
• Once the frozen trace buffers are stored into the hdcp.tgz dump file they are unfrozen and resuming tracing.
• There are also several more detailed logs included in the collection.
• Analyzing these traces and logs may provide Hitachi with the reasons or clues to why IO maybe having problems.
• It is possible we need multiple Mode31 dumps to capture a non-persistent issue.
• The Mode31 dump can be much larger than a detailed dump without Mode31 so can take a longer time to produce and transfer.
• Enabling SOM31 does not usually impact performance.
  • Experience shows that impact has been observable when storage is already overloaded or running at maximum stressed load and even so the overhead has been relatively small.
  • If you are concerned your Storage and host performance can be monitored and SOM31 is dynamic and can be enabled/disabled with immediate effect.
• In some cases you may be requested to trigger a Mode31 dump for SOM76 consideration.  See: How to Determine When to Use System Option Mode 76