1

I have a VPS with a hosting provider. Recently, I discover that my load average is just extremely high. I have a 2 core VPS, and the load average can go up to 9 or 10, as shown below!

enter image description here

I initially thought that this was because my CPU/RAM not enough, however, the VPS provider told me that it could be that my IO load is very heavy ( or other VPS on the same dedicated machine's IO load) is heavy.

I have no way to verify the claim.

Anyway I can check the usage of IO? And can I know whether the high IO usage is due to my VPS, or other VPS cohabitat on the same physical machine?

Graviton
  • 2,865
  • 12
  • 42
  • 64
  • You can use `vmstat` to find out how much I/O your `vm` is doing. Sadly, OpenVZ doesnt' support `iostat` or `iotop`. If your I/O is high, I'd accept the host's explanation. If not, then you are a victim of I/O overload coming from other containers. – David Schwartz Apr 10 '13 at 03:37
  • David, this is the problem here: how can I be sure that it is some contamination from other containers, or my IO load is overflowing me? – Graviton Apr 10 '13 at 04:24
  • What difference does it make to you whether the problem is contamination from other containers? What you need to figure out is whether the load you are seeing is justified based on the I/O you are doing. If so, then it's your problem. If not, then it's your host's problem. It's the host's job to figure out *why* you are seeing a problem you aren't causing, not yours. You just need to figure out if *you* are causing the problem or not. – David Schwartz Apr 10 '13 at 19:07
  • @DavidSchwartz, how can I figure out is whether the load I see is justified based on the I/O I am doing? I perform one simple database write query, the load average doesn't change much, but it takes a long time for the write query to execute, and around that time, I am not doing any other IO. – Graviton Apr 11 '13 at 02:39
  • As I suggested, by using tools like `vmstat` to see how much I/O you are doing. If you're doing a lot of I/O, you shouldn't be surprised that it's slow. If you're not, then you have a legitimate complaint to bring up with your hosting provider. – David Schwartz Apr 11 '13 at 02:40
  • Yes, I did some [vmstat test](http://serverfault.com/questions/497949/high-load-average-low-cpu-and-memory-usage-and-minimum-wait-time-for-io-operati), and I think my IO is quite OK. – Graviton Apr 11 '13 at 03:26
  • Then you should complain to your provider that you have ruled out your I/O as the source of the problem and that their server cannot provide sufficient I/O throughput for the containers running on it. – David Schwartz Apr 11 '13 at 03:29
  • I am not sure how to argue with them, I think I have done everything necessary, see [here](http://serverfault.com/questions/498253/dd-test-indicated-that-it-took-51-second-to-transfer-1-byte-vps-hosting-provide) – Graviton Apr 11 '13 at 03:35
  • 2
    Don't argue. If they don't fix the problem, stop doing business with them. – David Schwartz Apr 11 '13 at 03:36

1 Answers1

1

Looking at the screen-shot you posted I can see that your VPS is on OpenVZ container based virtualization. Unfortunately,there is no way for you check IO stats due to how container based virtualization works. Your host is correct that the IO load will reflect in your load average,however this seems strange. Normally disk IO isn't what pushes you to a high load average,but CPU access or load causes the problem.

I would postulate that your host is giving you a bogus answer to cover up massive over-utilization/overselling or abuse of the node by another client. Your host should be able to give you a better idea on if the high load is caused by Disk IO or CPU , and if they can't I'd suggest finding a new provider. ( I work for a web host and manage OpenVZ nodes as my day job.)

Jacob
  • 9,204
  • 4
  • 45
  • 56