0

whenever I use rm or cp on a bigger file structure the disk i/o is going up to the top, but not with the focus on cp or rm, it is the software raid sync. My knowledge in this area is limited. The topics I found were about finding the source of it and nothing further. As I know it already here the question.

  • Is this sort of behavior normal for a system raid?
  • What can I do if I/O nice c3 is changing nothing due to the burst from the raid?

enter image description here .

md2 is mounted with nobarrier

Drives: HGST HUS724020ALA640

# cat /proc/mdstat
Personalities : [linear] [multipath] [raid0] [raid1] [raid6] [raid5] [raid4] [raid10]
md1 : active raid1 sdb2[1] sda2[0]
      523968 blocks super 1.2 [2/2] [UU]

md2 : active raid1 sdb3[1] sda3[0]
      1936077760 blocks super 1.2 [2/2] [UU]

md0 : active raid1 sdb1[1] sda1[0]
      16768896 blocks super 1.2 [2/2] [UU]

unused devices: <none>

1 Answer 1

1

I would assume this as normal behaviour.

As your rm process doesn't need much for iterating through the directory structure and delete files, for the underlying disks it causes more stress. This creates iowait and leads to this situation.

From your screenshot, i assume, that 37 MB/s write on a big directory structure (which means many IOPS) is quite fair for a 2 spinning-disk-array. As long as your throughput values regarding linear write and iops are ok, theres nothing to worry from my point of view.

For the not trivial task of limiting io see: How to Throttle per process I/O to a max limit? ionice is a approach, but will only become effective, if there is concurrent workload.

11
  • alright thanks. Do you have any idea/suggestion on what I could do to lower the usage. I don't mind if it takes 3 times as long. As stated above ionice failed. :( Feb 10, 2015 at 16:28
  • This isn't a trivial question: See unix.stackexchange.com/questions/48138/… for more details.
    – Henrik
    Feb 10, 2015 at 16:35
  • Anyway you should test your raid with dd, fio, bonnie... to see if it outputs, what you can expect from your setup, to totally exclude a misbehaviour.
    – Henrik
    Feb 10, 2015 at 16:39
  • I edited my answer with the additional information on throttling. Thanks for the accept/upvote!
    – Henrik
    Feb 10, 2015 at 16:44
  • Quick question. Does that mean ionice would slow it down once the jdb2 is going up. From what I noticed it doesn't slow down at all compared to rsync which is running before with c3 and being slowed down to tops 60% disk i/o usage. (98% without ionice setting) Feb 10, 2015 at 16:49

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .