Can someone please point me to documentation about what happens if either of these can't send data out to Dynatrace:
I assume they each have some sort of persistent queue so if communication is lost it will collect the data and then send on reconnection - but is that right?
And what limits are in place as to how long it can accumulate data to send before running out of disk/memory?
I did have a look but I am not used to Dynatrace docs yet so a pointer would be greatly appreciated.
Hi @Keith_Harding ,
Please note that Dynatrace OneAgent doesn't cache the data to avoid making utilization issues on the monitored server, that's why it's recommended to use the ActiveGate and to be installed closest to OneAgent with fewer/no firewall points between them, as the ActiveGate bundles OneAgent Traffic and is not limited to that, you can find more details here regarding ActiveGate purposes and functionality. and check ActiveGate acts as a secure proxy as well
also, I think it would be better as well to check and understand the Network Zones
Dynatrace OneAgent does cache data. I say this from experience, and even yesterday I had one case where I had infrastructure data for several hours, before the agent connectivity was back up. The application modules/subagents seam to have a different behavior, also from empirical observations.
I'm not sure if the data that @Keith_Harding references is still valid, because it's some years old, but would love to know more details.
Thanks Antonio, thats great real world info.
It would certainly be a good to see an official answer. Some of the info I saw was "we don't cache because we don't want to use to much resource on the machine" which makes sense but as you say in some cases it does cache.
I also heard that they may drop metric granularity as the time unable to connect grows - that makes sense.
You pont about different behaviour per module makes sense. Caching some metrics is probably low disk need but caching traces and logs could require a lot of disk evry quickly.
Thanks for your thoughts.
Thanks Mohamed, much appreciated.
For anyone else looking I also found out:
Dynatrace also has automated monitoring and alerting of conditions where large numbers of OneAgents stop sending.
And this is a previously similar question: