We recently had a service which failed in WebLogic for a domain running on 3 nodes. Apparently during one of our maintenance windows in Dynatrace when WebLogic patching took place, the WebLogic servers were brought back up before our maintenance window expired which is fine. However, when the WebLogic nodes were brought back up, a specific service failed on 2 of the 3 nodes. There was no sign of this within Dynatrace which I understand why.
Because this is a critical service to our business, we started looking for a way to have Dynatrace monitor that the service is running on all 3 nodes as you can do with progress groups. However, I can find no method that would allow me to for example say if there are not 3 instances of this service, raise a problem.
I would think there would be an option like "Availability monitoring" for services like there is for process groups as seen below. As you can see, we do use this for the process group (WebLogic domain) which is also where the service that failed on 2 nodes lives.
The above however is only for the process groups (WebLogic domains) themselves, and not the services under it. If I look at the settings for the service, there are no options for availability monitoring.
As you can see below, there are 3 nodes with 8 services.
One of the services which is called rocServiceWeb should be running on all 3 nodes at all times.
Is it possible for Dynatrace to be configured to look at a service and ensure there are X number of instances running and if not, raise a problem like you can do with process groups?
I spend time going back through documentation and looking through the configuration API, but I could find nothing that really pointed me to a solution for an instance count and I am not sure one even exists which I find odd considering there is one for process groups.
Below are the properties of the service.
Feedback is welcomed.
Solved! Go to Solution.
From my perspective alerting would be extremely simple and useful as you describe it and we have all the functionality in place already (except the service instance count metric, which I also can't find.)
I will contact the responsible PM and ask why we did not export the service instance count metrics.
I spoke with the responsible PM and he told me that he will put the exposure of service instance count metric onto his roadmap. Once that metric is exposed, you can then use the normal metric alerting to raise any kind of events and trigger Davis.
Is there any timeline on this? We just finished migrating from AppMon, where this sort of task was very easy, to the newer Dynatrace product and were surprised (and disappointed) to find that this capability was not already in the solution.
The ability to be alerted when a service was being improperly load balanced is something that had earned its predecessor several wins with our application teams.
That we are now having to go back to those same teams and tell them it is not a possibility yet has diminished some faith in the tool, making it more difficult for us to drive a successful adoption of the new product.