junkins
By:
junkins

Disk I/O being overloaded

November 21, 2014 13.4k views

I am using Zabbix to monitor my droplet and I am now getting warnings fairly regularly of the Disk I/O being overloaded which, in my Zabbix configuration, is triggered by the cpu.util.iowait metric being over 20%. Is there something I can tune in my droplet to make the I/O more efficient or is this a matter of the load on the host server due to the other droplets hosted there? The droplet is running CentOS.

3 comments
  • I get absolutely same error

  • Over the last few days I'm seeing this same behaviour on my droplet... could you share how did you solved this?

    The strange thing it's that iotop it's reporting almost zero usage, but NewRelic informs about 20% I/O wait with 100% spikes when doing stuff like updating packages, rsync, etc.

  • Hello All... since April, 1th I´m experiencing the same... High CPU IOWait Time...
    How did you solved? Or any tip?

2 Answers

In most cases either one of two things is occurring here:

  • A process or processes on your droplet are doing a whole lot of disk i/o operations. You can determine if this is the case and seek to optimize the service.

  • Your droplet's hypervisor is experiencing degraded performance due to another droplet on the same physical hypervisor. We've greatly improved our monitoring and other backend services since this question was first posted making this less likely but should you find this is the case you should open a ticket with our support team so they can investigate and resolve the issue.

Have another answer? Share your knowledge.