Good morning! Yeah I have been side lined working on more end of the year stuff. I was trying to remember, are you building a OneAgent plugin or ActiveGate plugin?
Asking because when I tried to go the OneAgent route, I found no method to get get to a level of where each queue name could be tagged. This is a requirement for us as that is how we assign ownership. By going the route of the ActiveGate plugin, that allows me to create custom devices so I take each queue and do exactly that, I turn each one into a custom device. This then treats each one independently and allows for auto-tagging. I also placed them under their own technology called "EMS Queues" so that they are separated from the core "Tibco EMS" technology. The reason for this is because there is one team which owns the support of Tibco EMS itself for things such as upgrades, patching, etc. and then the queues are owned by multiple teams.
We are bringing in....
- Queue Depth
- Queue Size
The reason for this is due to the amount of custom metrics which each metric consumes. Because we go to a depth of the individual queues, so it adds up fast depending on how many queues you have.
I too am still playing around with the aggregate. I will be honest, this is one area of Dynatrace that always seems to confuse me that I can never keep straight in my head and often just have to play with it. SUM, COUNT, MAX, etc... I always have to go back and look up exactly what each one means.
There are pros and cons to each method.
OneAgent Plugin
Pros
- AI is utilized.
- The relation ship between the said queues, services, processes, etc. are understood by Dynatrace.
- Does not use as many custom metrics as the ActiveGate version due to higher up the tree scope.
Cons
- All queues are placed under the technology "Tibco EMS" process.
- Queues are not treated independently to a point that they can individually be tagged manually or through rules.
- OneAgent plugins do not have the ability to create custom devices.
- Does not scope deep enough down to each independent queue level.
ActiveGate Plugin
Pros
- Has the ability to create custom devices, therefore allowing each queue to be turned into its own individual device.
- Haas the ability to scope down to each individual queue.
- Because queues can be individual devices, you can manual and / or auto-tag like you would anything else.
- Queues can be separated out from under the "Tibco EMS" technology group which is a huge benefit if you have multiple teams supporting different things within EMS.
- Each team can create custom thresholds under "Custom events for alerting" and can be very specific due to the fact that the plugin uses the queue names as a dimension.
Cons
- The AI is not utilized.
- The relationships between said queues, services, processes, etc. are not understood or shown by Dynatrace. This one might not stay a con as I think it's due to the IP being assigned more than once. Working on this one.
- SmartScape shows these custom devices which are queues as hosts.
- Can utilize a large amount of custom metrics.
I am still working on the ActiveGate plugin and I might be able to reach a point where the relationship between the queue and services is shown. I was talking to some folks at Dynatrace and I think it's due to the fact that the same IP is being used which is why it does not track. So currently I am working on the structure a bit to see if I can get around that.
Both options are a trade off from what I have found. It really depends on the need at the end of the day. For us, we must be able to obtain these metrics, but also assign ownership to each queue and therefore custom devices are a must.
While I can not speak for Dynatrace, I have had a few conversations with these great minds over there about this. The out of the box solution for EMS and any other messaging service for that matter would be for Dynatrace to just go one level deeper. As you can see below, Dynatrace out of the box detects the EMS queue names. It already see's these queues. If you click on them however they just take you to the SmartScape and do not go any deeper. My hope is that Dynatrace will enrich the product more for messaging services so that not only do you see the queues, but can click into them and then see all of the metrics which are vital to customers. Such as queue depth, size, etc. This is a challenge as every messaging service is not the same. For example, when it comes to Tibco EMS, they would need to use the tibemsadmin. For something like Pub/Sub in GCP, they would most likely need to tap into StackDriver. There is a big challenge there, but I think it's one they need to meet due to the fact that if you can not get down to these levels of metrics when it comes to messaging services, it becomes somewhat useless. I do think we will see headway on this from Dynatrace eventually, but for which messaging services and when - Anyones guess.
In terms of the process group detection, are all the servers running the exact same version of EMS? If so, I would bet that it's the different paths thats making Dynatrace separate them which I would expect. That being said, if you go to "Settings", "Processes and containers", and select "Process group detection", you can build a new "detection rule" there which should then allow you to group them up.
Hope that helps! I will release what I have soon. I need to get over to GitHub.
Thanks!
Featured Posts
