Best practice in Agent Group definition?
Hi,
We have a web application with two tiers in the presentation layer(web layer). these two tiers are in same technology but running different code. I would like to know the best way to define agent groups for this application. Should we define one agent groups for each tier or define one agent group for entire web layer and apply sensors for both tier in one agent group?
Thanks,
Sreerag
Hi
Agent Groups should be considered like Application Tiers. In your case it seems that these two agents actually belong to the same tier (Frontend) - so it would make sense to create one agent group and use sensors that will work for both agents.
The AGent Groups also play a big role in the Transaction Flow - which is the visual representation from your PurePaths.
If your two Frontend Agents "talk" with each other then I assume you also want to see the transaction flow showing the transactions going into each agent - in this case I recommend two creating Agent Groups.
Andi
Hi Andi,
The two sections of the frontend doesn't talk to each other but we have around 40 servers in the frontend (half of them running different code than the other half). Would you recommend to put the agents from all these servers into one Agent Group? Will there be any issues in applying custom sensors when we group two frontend tiers into one. Some sensors won't make any sense to half of the agents in the agent group.
Also would like to know if there is any benefit in using less number or Agent Groups than creating one agent group for each code type so that agents from all server running the same code belong to one agent group. Is there a limit in the Agent Groups in DynaTrace for a profiles or is there a recommended number. Does having more agent groups impact the DT system performance?
Thanks,
Sreerag
Sreerag,
As Andi mentions, you should look at your app from a functional point of view (looking at application "Tiers") and break it into agent groups along those lines. There are a couple of reasons for that. One, as you mention, is that sensors are placed and configured at the agent group level.. Second, the agent groups define the view of the Transaction Flow. Surely you would like to differentiate the time spent in one FE tier from the other FE tier, and in the time spent communicating from each of those tiers to the backend. This will happen if you make each separate frontend tier its own agent group. Another benefit is that you can set filters to just look at transactions that pass through a particular tier. Bottom line is that I recommend you consider having two FE agent groups based on your description.
There is no performance penalty with having multiple agent groups. The only downside (and this is changing) is that it is sometimes tedious when you have settings that you want to apply to all agent groups. In those cases you would need to make the changes in each group. Small price to pay for the benefits listed above.
Thanks for the recommendations Rob and Andi. Looks like defining Agent Groups for each tier is the best way to go.
Hello,
Thoughts on the similar lines...
Is it a best practice to even split each of the outbound calls from a particular tier as a separate Agent Group?
Idea is to visually see the complete end-to-end application flow (using Transaction Flow) to determine how much time is spent within the tier and the outbound calls separately from within that tier.
Thanks,
Abhijeet G
