Microsoft – Windows 2008 – R2 / QLogic HBA (QLE2562) – Drivers (2012-03)

Using the new “Activity Monitor” that comes with “Microsoft SQL Server” – v 2008/R2, we have being able to lock in to the fact that our biggest wait category is “Buffer IO”.

So what that to do when faced with the following Perfmon metrics:

  1. Queue Length (High – over 20)
    • Logical Disk – Avg. Disk Queue Length (High – Over 20)
    • Logical Disk – Avg. Disk Read Queue Length (High – Over 20)
  2. Queue Length (Avg. Disk Sec > 15 ms)
    • Logical Disk – Avg. Disk sec/Read ( Greater than 15 ms)
    • Logical Disk – Avg. Disk sec/Write ( Greater than 15 ms)
  3. Disk idle Time (less than 10%)
  4. Disk Read/Write Sec (Lower than the rated numbers)
  5. Logical Read
    • Logical Disk – Disk Reads/sec
    • Logical Disk – Disk Writes/sec
    • Logical Disk – Disk Writes/sec

Well, keep looking and doing.  And, talking.

One of the things I did was talk to our Storage Engineers.

And, then talked to an Industry Luminary.  Spoke to my Uncle Charles.  He said it appeared that we have enough “firepower” and rather than add on more fire-power may be we should just do a bit more instrumentation (that is get more insight into what the problems were).

Our system is quite complex and we are just not sure what the expected metrics really are.  And, so what we ended up doing is taking the easiest path.

Apply patches.

  • Microsoft – SQL Server – 2008 – R2 – SP1
  • NetApp – Data OnTap DSM – v 3.5
  • NetApp – Fiber Channel (FC) – Host Utilities (v 5.3 to v 6.0)
  • QLogic – HBA – QLE2562  – v 9.1.9.25 to 9.1.9.47
  • HP – Proliant – SmartArray – P4100 Controller –  v 6.22.0.64

 

 

References

2 thoughts on “Microsoft – Windows 2008 – R2 / QLogic HBA (QLE2562) – Drivers (2012-03)

    • Yes, the patching helped.

      1) I will advise for you to take a good look at “SQL Server”\ “Management Studio” \ “Activity Monitor” and
      see what your sustained I/O Metrics – In the Overview section, under “Database I/O”

      2) Also, look at recent expensive queries and order by “Physical reads/sec”, “Logical Writes/sec” , “Logical Read/sec”

      3) Add memory to the DB – Plenty of RAM and hopefully you will not go to disk as much

      4) Add Flash Card to Storage Sub System (in our case NetApp)

      5) Adjust Queue Depth

      6) Determine your Storage IOPS (https://danieladeniji.wordpress.com/2012/07/05/qlogic-determine-iops-using-qconvergeconsole-cli-version-1-xxx/)

      G(ood) luck,

      Daniel

Leave a Reply

Please log in using one of these methods to post your comment:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s