Searching for QFull and Queue Depth

[Gabrie1]

Hi,

In discussions over tackling "Lost access to volume" messages and "Performance deteriorated" messages, we're trying to find the cause of these. One thing we're looking at is to change the queue depth settings, but before I want to change those, I first want proof that we really need to change these. I can find a lot of posts and KBs on how to set the queue depth, but there is little info on how to check the queue depth reached and if it is even necessary to change the queue depth.

Is it correct that to search for QFull messages I should search Log Insight for the following string: "D:0x28" ?

[kastlr · Accepted solution]

Hi Gabrie,

I assume that's not an easy task, as depending on the used arrays the behavior might differ.

To the best of my knowledge (at least) some arrays only report QFULL messages when there controllers are totally utilized.

In any case it would be better to focus on the array monitoring tool to figure out if a QFULL condition occurs.

Based on my personal experience, the majority of issues with "Lost access to volume" is caused by ATS Hearbeats which where introduced by VMware in ESXi 5.5 U2 and was seen with multiple arrays from multiple storage vendors.

Disabling that feature and reverting back to classic VMFS heartbeats usually stabilizes the environment.

ESXi host loses connectivity to a VMFS3 and VMFS5 datastore (2113956)

Regards,

Ralf