OneAgent manual management

[Babar_Qayyum]

Dear All,

Do we have a way to manage the agents from one endpoint to another endpoint manually?

e.g. agents are connected to 2 x security gateways and now I want to upgrade one of the gateway or the server but before that I should be able to move all the connected agents to another security gateway for doing the scheduled activity.

Regards,

Babar

[Julius_Loman · Accepted solution]

Agents will move to another valid endpoint (server url) automatically. Agents also continuously check each endpoint in the endpoint list, so you don't have to "redirect them". They should fail over quite fast.

[Babar_Qayyum]

Hello @Julius L.

Is there any chance to drop some performance data while fast failover?

I am asking because it can trigger the configured alerts.

Regards,

Babar

[Julius_Loman]

I did a quick check in my lab and the failover took approximately 40 seconds. No data was lost (agents have buffer for data).

Just a quick note - there are multiple connections from monitored host to the gateway. Each instrumented app has its own one, os agent, plugin engine and log analytics also have individual connections. Its normal to see that agent parts or instrumented applications report to different endpoints at the same time.

[Babar_Qayyum]

Hello @Julius L.

Thank you for the exercise and providing me the facts. Just one more clarification about the buffer.

In case failover took more time then how long agents will keep data.

Regards,

Babar

[Julius_Loman]

Buffers are sized in bytes. So the duration how long agents keep data depend on how much data the agent sends.
See this thread https://community.dynatrace.com/questions/200682/dynatrace-managed-disaster-recovery.html

[Babar_Qayyum]

Hello @Julius L.

It means if we provide much disk to the agent then the survival time can be increase in case of unavailability of security gateway.

Is my understand correct?

Regards,

Babar

[Julius_Loman]

Based on my observation, the data is kept in memory. If you shutdown the agent or deeply instrumented process without connection to cluster or gateway, the data is lost.

[Babar_Qayyum]

Hello @Julius L.

Do you mean the agent will buffer the data in the host memory and it depends on the size of the data to buffer before drop/lost the data in case no connection to the cluster or gateway?

Regards,

Babar

[Julius_Loman]

Correct. But this is strictly based on my experience and observation.

[Babar_Qayyum]

Hello @Julius L.

Fair Enough !!!

Maybe someone else can
also share the personal experience with us for more understanding.

Regards,

Babar

[Christoph_Neum]

Yes, the agent buffers data only in memory, not on disk.

The failover should happen within 10 seconds + the time it takes to establish a new TCP connection (it happens on every heartbeat, which has a 10 sec interval).

[Joe_Hoffman]

Babar, Julius's answers seem to address your questions. Can you clarify what additional understanding you are needing? Be sure to explain your use case so we understand your context of questions.

[Babar_Qayyum]

Hello @Joseph M. H.

Please consider my apologies in advance in case I am asking something weird.

Basically I am trying to collect the maximum technical information to complete the TDD.

Do we have an upper memory settings for the agent to buffer the data or it will use the host memory without any limitation?

Regards,

Babar

[Julius_Loman]

Yes, there is a limit. Actually you can notice it in the logs. However, you cannot change it.

[Babar_Qayyum]

Hello @Julius L.

Thank you for sharing this information.

May I know that how much did you notice till now?

Regards,

Babar