Detect files behind I/O wait issue

We are running Asterisk 1.8.22.0 which uses NFS for sound files (for IVR’s), voice mails etc. Every so often the load average gets real high. Anytime that is we see the I/O wait in top go over 7%. The longer it stays up there the higher the load climbs (sometimes all the way to 120%). This is all very understandable and expected (high I/O == low CPU idle which translates to high load average). The “episodes” episodes of high I/O don’t usually last two long (10-30 seconds max) but when they do hurt call quality. It seems almost as if it works fine till the box reaches a tipping point and can’t cope with the I/O wait. When the issue is ongoing iostat is showing the disks are behaving fine which leads us to believe it’s an issue with files accessed over NFS. This is only a problem on our CentOS7 boxes. We have the same version of Asterisk on CentOS6 boxes with no such issues. I assume it has to do with how the box mounts the NFS drive? On both boxes in /etc/fstab we have

xx.xx.xx.xx:/var/lib/my_mount /var/lib/my_mount      nfs soft,bg,intr,noatime 0 0

On CentOS6 we have installed

nfs-utils-1.2.3-64.el6.x86_64

On CentOS7 we have installed:

nfs-utils-1.3.0-0.33.el7.x86_64

Any pointers?

Related:


Leave a Reply