My question is simple. I just can't seem to find much information on it googling around. If I set up Windows Error Reporting to create memory dumps when an IIS app pool crashes (See article here for rundown.) could it cause any noticeable degradation in performance in terms of IIS serving up apps and websites?
We are looking to set this up in production to assist in tracking down issues when an app pool crashes. Also, is something like this recommended for production servers?
I assume you're using the LocalDumps Registry key to achieve this.
That Registry key has a setting called DumpType where you can specify the amount of data being collected.
The time it takes to capture the crash dump will mainly depend on the amount of data to be written to disk. A full memory crash dump of IIS can easily take 4 GB, which may take 8 seconds to be written to disk at a 500 MB/s disk throughput rate.
During those 8 seconds, another instance of IIS may still serve pages from cached files very fast, but might have trouble serving files that need to be read from disk.
You could mitigate this a bit by spending an extra disk or different partition for the crash dumps. You'll then just have the memory, CPU and SATA overhead.
Anyway, you'll not leave this setting on for very long. Just capture a few crash dumps and then disable it. You'll make your customers happier if you resolve the crash. Therefore, the performance impact IMHO is acceptable.
If you want to know the exact impact, you'd need to set up a load test system, serve pages and implement a website that crashes. You'll then notice (or not) the performance degradation and you'll be able to measure it.