I have an odd issue ever since I added 3 new drives in my system on a 1x PCIe HighPoint RocketRaid 2300 controller.  Every time I do any reads on any of the drives, the read count jumps 3 - 5 times what it would on any other drive.  Below is a Parity check I just did that shows the issue:

 

 

I had been using the system for a few hours before starting the parity check and did not clear statistics before hand, but none of the read counts on any of the drives were above 100k.  After the parity check, Drives 5-7 are all way higher than the rest by 3-4 times.  Further, read counts seem to edge down by 4 million reads from Drive 5 (21), Drive 6 (17), Drive 7 (13).

 

Parity and Drives 1 - 4 are all on the onboard SATA controller, while Drives 5 - 7 are on the RocketRaid 2300.  This doesn't really bug me, but the parity checks are noticeably slower (45 MB/s versus 60-70 MB/s) and reading from these drives seems a little slower.  What would cause this, or is this normal?  Is this just a driver problem for the RR2300?

 

I doubt anyone will be able to answer these questions, I'm more curious if others are seeing similar scenario's and could this be affecting performance of others' systems?

 

I am running unRAID 4.4.2.  Motherboard is a Gigabyte 965P-DS3 with 6 on board SATA connectors.

 

Click on 'Clear Statistics' and then start a parity sync - let it run for say 30 seconds & then cancel, then report back the read counter values.

 

Before the 'cache disk' feature, the webGui reported the read/write counters kept track by the unRAID driver - this would count the number of 4K block read/writes generated to the disk drivers.  After 'cache disk' feature was implemented, since cache disk is not controlled by unRAID driver, a different method for reporting these counts is used.  That is, directly reading the disk driver I/O counters.  Sometimes this leads to "strange" results because often disk drivers will "combine" multiple sequential commands and actually issue just one command to the disk.  Whether the driver reports this combined I/O as 1 I/O or increments the counter according to input command count is up to the driver.

 

Probably what's happening is the RocketRaid driver is just reporting I/O counts differently - not surprising.

 

As for your slow down - that's also not surprising and is probably due to raising count of array drives from 5 to 8.  Could also be the fault of the RocketRaid card/driver - it's really hard to determine that unless you start experimenting with different hardware.

Here is a screenshot of the Parity check running for just 30 seconds.  Looks to be the same ratio of reads for each drive to me.

 

 

Obviously not hugely important.  I would rather see a 64-bit build or dual Parity before this is fixed