Large Sleep counts causing THREADPOOL wait?

by Chris Woods   Last Updated March 14, 2019 20:06 PM - source

When I run sp_whoisactive throughout the day, I usually see we have around 500-700 connections on our server. We are occasionally getting threadpool waits and high signal wait ratios (so we are thread starved at times). However, of these 500-700 connections only 10-25 of them are active. I wasn't sure if a sleeping spid is still bound to a thread while it sleeps? Could that be causing our high signal wait ratio and threadpool waits? I have a feeling our 3rd party application is not properly closing out connections and is causing us issues, but wanted to check first before I bring this up with them. Thanks!

Answers 1

I suppose they could be bound to a thread while they are inactive, but they shouldn't be causing threadpool waits if they are just sitting there. I'd look at what those active sessions are doing. See if they have a "CXPACKET" wait type, that could show that those active sessions are using multiple threads to perform their actions. Here is a good article about CXPACKET with some troubleshooting tips:

Nabil Becker
Nabil Becker
November 05, 2015 14:26 PM

Related Questions

Nightly EDW Build Wait Stats

Updated March 21, 2017 12:06 PM

Troubleshooting Preemptive_Xe_Dispatcher wait type

Updated December 08, 2016 08:02 AM

Heavily used SQL Server with only a few wait types

Updated January 25, 2018 10:06 AM

CXPACKET wait on SQL Server

Updated January 24, 2019 02:06 AM


Updated June 05, 2015 11:02 AM