CPU I/O wait + linked disks

[svendb1803]

Hello

I am looking for a way to link the metric CPU I/O Wait time (builtin:host.cpu.iowait) with the disks the CPU is waiting for in the data explorer.
Anyone has an idea how to construct such a query?

thank you in advance

[Sam_Richardson · Accepted solution]

Hi @svendb1803, great question.

CPU I/O Wait time is a great metric for identifying hosts that are experiencing wait time generally, however you may find Disk Average Queue Length (builtin:host.disk.queueLength) to be more useful for your purposes. This metric can display both the host, and disks, in the same view.

Unfortunately, this metric doesn't show definitively whether a disk is causing wait time or not, but it does show disk queue length. If you already know a host CPU that is experiencing wait time, filtering this query by host should shed some light on which disks are working the hardest.

Here is the Advanced Mode query I created:

builtin:host.disk.queueLength:splitBy("dt.entity.host","dt.entity.disk"):sort(value(auto,descending)):limit(20)

I've also attached an image of the visual query.

I hope this helps!