On my Debian VPS, I am getting peaks of IO wait, which blocks my server for a few minutes. On the screenshot, you can see the Idle (green) which suddenly drops for IOwait (pink), which eventually blocks even the logging process (white stripe).
There is nothing strange happening on the machine (as you can see, most of the time, the machine is waiting) and the system resources are free (800 MB of RAM is unused).
I am worried that this is actually a problem of my VPS provider, but I don't know how to prove that.
Is there a way to find which can be causing this?
Thanks!
EDIT:
What can be the other small peaks, which are repeating probably every 5 minutes? How can I find how they are caused?
/var/log
during the hangs? Does it coincide with andcron
jobs? Installiotop
and runiotop -Pbto > /tmp/io.log
. Leave it running through one of those hangs and check the file to see heavy users. Please add any items of interest to your original post.iotop
command now and leave it running. Wait until one of those spikes and then you can look back at the file/tmp/io.log
and match up the timestamps with the period you're interested in. You should be able to see high I/O usage processes, if there are any.