Showing results for 
Show  only  | Search instead for 
Did you mean: 

Can we monitor number of active connections on a websphere server using DC RUM?


We have websphere Application hosted on windows server where different TCP/IP ports were configured from application end to receive data from different sources.
The ports at WebSphere is configured for 100 connections/port, after that it does't allow new connections. Is it possible to get count of connections using DC RUM monitoring so that an alert is triggered in advance when connection count reaches close to maximum e.g. an alert is triggered when connection count is 80



Hi Rattan,

You can refer to this question on the forum as its very similar to your question -

In short, it would be very difficult to measure active connections in DCRUM, but you can measure connection attempts, connection errors and idle sessions (open, but with no active traffic).

Dynatrace Pro
Dynatrace Pro


This is true that measuring number of active connections would be challenging.

I have.a further question though: how exactly would you define number of active connections?

Connection is opened by client, but either party can decide to close the connection. In HTTP, typically server decides when to close the connection, and not on the TCP level, but rather HTTP, by sending the "close" header, which then prompts the client to send FIN. Server may also close connection right after sending response to the client (though in http/1.1 that's counterproductive). Normal HTTP operation will maintain connections open to speed up response to the client. And so on. There are many paths to follow.

So my question is how to account for "active" connections?

Is it about TCP sessions that are truly sending data in parallel, is it about connections which are just open from the TCP state machine perspective (but may be idle from the network transmission perspective - please note that server or client can choose to close such connections at any time), or is it about connections which are active from the server workload perspective (i.e. client sent a request and now waits for server's response)? There are many paths to follow.

I'd like to understand what kind of calculation of active sessions would make best sense.