7

I do a lot of work with my Linode that currently hosts a site, but every now and then (usually once every two months or so) - I'll get a warning about my IO rate being too high (Usually just higher than 6000). The last one I got said: "Your Linode ... has exceeded the notification threshold (1000) for disk io rate by averaging 6557.69 for the last 2 hours".

I'm a bit worried about this, but don't really know what to think. Is it healthy? Looking at my server graphs I never see anything special, below are two 'regular' days for my low traffic site (note that the two major spikes are my rsyncing to backup the server). Also note that even though I rsynced twice that day, I didn't get a warning come through about those. The graphs are as follows:

http://i.imgur.com/vLo7A.png

http://i.imgur.com/OIHT9.png

I've tried looking at iotop but whenever I take a look everything looks healthy.

Any ideas?

  • Where did you get this warning _"Your Linode ... has exceeded the notification threshold..."_? – quanta Nov 22 '11 at 08:53
  • @quanta It was emailed to me via some automatic thing wired up to my Linode. –  Nov 22 '11 at 15:18
  • 6000 ops/sec does seem high if you don't know what's causing it. Do you have daily backup? It might be that, or some other batch operation. – UpTheCreek Nov 24 '14 at 14:36

2 Answers2

5

6557 per hour means less than two disk operations per second. 6557 per two hours means less than one per second. Your only problem seems to be overly sensitive notification threshold.

I would investigate if I've got 50.000 per hour or so (but this still would not constitute a real problem).

kubanczyk
  • 13,812
  • 5
  • 41
  • 55
  • Whow. Did not even see that maths. Yeah - one IO per second is a non-working server. Heck, doing nothing it may do more IO due to logging. – TomTom Dec 27 '13 at 08:06
  • 1
    It's not 6557 operations in two hours, it's an average of 6557 ops *per second* over the last two hours. From the Linode manager: `Average Disk IO ops/sec over 2 hours exceeding this value triggers this alert.` – UpTheCreek Nov 24 '14 at 14:32
0

Disk IO are writes and reads to/from disk. I know nothing about your stack, but a common cause for spikes that don't correspond to traffic are log rotations/compression.

r_2
  • 335
  • 3
  • 9
  • Except the spike is no spike given how the answer takes the math apart into funny areas - less than one IO per second.... – TomTom Dec 27 '13 at 08:07