I have some synthetic measurements that were occasionally giving some DNS errors. These measurements were being made from a private synthetic location running on a Linux server.
I couldn't imagine why they were occurring, so decided to dig further. As it was a private location, I just dumped all the DNS traffic.
To my surprise, there was the DNS query packet, than there was a second one after a little bit more than 5 seconds, and than a correct reply.
So, I imagined something was wrong with Dynatrace, but I believe I quickly found the reason for this behaviour:
- Dynatrace times out DNS resolution at 5 seconds.
- Linux apparently retransmits DNS queries after 5 seconds...
So, what seams to be happening is that being the timeouts from Dynatrace and the OS (Operating System) equal, Dynatrace will always give a timeout before the OS retries. This wouldn't happen in Windows, as it apparently times out after 1 second.
There are two solutions for this:
- You change the default Linux behaviour, changing the DNS timeout to something lower.
- Dynatrace can give access to an "advanced mode" where DNS timeouts can be reconfigured.
Of course, I believe there might be additional impacts. I have also created a Product Idea, so that behaviour can be consistent when running namely HTTP monitors on different OS ActiveGates:
https://community.dynatrace.com/t5/Dynatrace-product-ideas/DNS-timeout-for-HTTP-Monitor-should-be-1-...
Antonio Sousa
