I'm trying to run some disk benchmarks to compare two identical drives in the array, one running XFS the other BTRFS.  The DiskSpeed docker doesn't appear to give me the results I'm looking for.  So instead I am running fio from the command line (it's in Nerd Pack), with the settings listed here:

fio --randrepeat=1 --ioengine=libaio --direct=1 --gtod_reduce=1 --name=test --bs=4k --iodepth=64 --readwrite=randrw --rwmixread=75 --size=4G --filename=/mnt/disk1/test/testfile

I notice the parity drive is active during this test - why is that?  I thought writes to /mnt/diskX (as opposed to /mnt/user) will bypass it, or am I mistaken?  Is there a way to bypass the parity drive temporarily (other than actually restarting the array without it selected)?

Edited December 27, 20205 yr by servidude

ALL writes to /mnt/diskX will always involve the parity disk.

 

Besides, your thought process is wrong.  If you were really going to purposely bypass the parity system (and there is a way to do it), then effectively you've impeded the system from being able to rebuild any given drive failure since the parity system on any given RAID solution has no concept of files being present, but rather the bits on each drive.  And creating a file (and then presumably subsequently deleting it) will amount to substantial changes to the contents of the drive.

6 hours ago, Squid said:

ALL writes to /mnt/diskX will always involve the parity disk.

 

Besides, your thought process is wrong.  If you were really going to purposely bypass the parity system (and there is a way to do it)

OK then, thanks for clearing that up.  What would you suggest as a better way for testing read/write speeds of individual drives?

Edited December 28, 20205 yr by servidude

You can use /dev/sd# if you don't care invalid the parity. ( When writing occur )