Network Latency for dynaTrace collector
Hello,
Can anybody please explain what is the minimum latency required for dynaTrace collector to listen from agents to collector.
I noticed in the document that it is necessary to run the collector service near to the machines where agent is installed, But we are running shortage of server for which I could dedicate one server to run this collector service near to the agent machines. I'm thinking to run collector server in the same dynaTrace machine which is in India. The server which I would like to monitor where Agent is installed is located in Sweden. Please suggest if this is feasible or if there will be any impact on dynaTrace machine where I am thinking collector service.
Thanks in advance
/Vasanth
Vasanth,
Agent in Sweden and collector in India is not a good solution. We like to see 2ms or less between the agent and collector. It starts to be a concern at 10ms.
Mainly this is due to startup times. During startup there is a conversation between the agent and collector for every class. if you have 30K classes, increasing latency from 2 ms to just 20 ms means startup will take 10 minutes. Also, the agent will evaluate latency at startup and if it is too much will decline to participate in instrumentation.
My advice is to find a way to get the collector in Sweden. If there is any additional capacity on the app server, even putting the collector there (not recommended!) is a better solution than what you are suggesting.
Rob
Rob, Thanks for the suggestion.
Hello Rob,
Is there any plugin from dynaTrace so that I can leverage network latency between collector server and agents where I have installed. If there any, i am planning to install it on collector server so that i could have a better network latency between collector and servers where i have installed agents.
/Vasanth
Hi Vasanth,
No, we don't have a plugin for measuring this. Normal network tools such as "ping" and "traceroute" should help here. Also: if you look in the dT startup log files you'll see that we do log some statistics for things such as "transformation time" which measures the latency for performing sensor injection. It will tell you if the transformation time is too long such that the agent will refuse to inject (for fear of slowing down app startup time too much).
Rob
Hello Rob,
Thanks for the suggestion!
I have checked the log folder C:\Program Files\dynaTrace\dynaTrace 5.0.0\log where I have installed dynaTrace, Also I am doing some testing since I need to change the collector server for our existing setup. In the log file I could not able to see logs for all time. I'm seeing logs for only our old collector server and dynaTrace seems to be not creating logs for our new collector server. Will there be something like there could not be any sink between my new collector and dynaTrace for logs?.
For my new collector I am able get all instrumentation details and all transaction details. Please see in below figures and let me know if this is healthy status if I can run with new collector server.
AgentStatus_newCollector.jpg, PurepathStatus_newController.jpg, Db_transaction_newController.jpg
Also, I always use to get this agent error for our both old collector and new collector Agent_error.jpg but it is not affecting any instrumentation details. Do I need to consider this error?..
Thanks in advance for your assistance.
/Vasanth
