Prospect asks if there is a way to monitor that there are sufficient number of NodeJS worker process (i.e. in the next screenshot is 1 of the 5 workers processes is down to get alert on that)
Thanks in advance
30 Oct 2017 09:31 AM - last edited on 08 Dec 2021 04:32 AM by MaciejNeumann
You can do this through Availability Settings for a process group.
If you set this to 'if any process becomes unavailable' then Dynatrace will open a problem if any of the worker processes goes offline or becomes unavailable.You could also set this to a minimum threshold. This is nice for scaling. If you know that in normal traffic levels you run at say 3 instances, but often bump up to 4 or 5 and don't want an alert UNLESS there is less than 3 running instances, you can set the minimum threshold to 3.
Hope this helps,
First thanks for your answer!
From what I understand and correct me please if I'm wrong the availability monitoring is to check if we have the minimum number of instances running the NodeJS and not the worker processes under each NodeJS instance.
In our case we have 4 NodeJS instances and each of them is running 5 worker processes and we want to be sure that all the 4 instances are running 5 worker processes...
Sorry about that, I slightly misread your question the first time through! Now I understand what you are speaking about - but, I don't believe that it is available within the product, at least yet. I will leave this opened and unanswered and let others from PM team respond.
Thanks for the willingness!
We will wait for the cavalry
06 Nov 2017 02:40 AM - last edited on 12 Apr 2021 05:25 AM by MaciejNeumann
I don'r fully understand the use case here. The number of worker processes is usually constant and defined through the cluster module or a process manager like PM2. Usually it's the number of CPUs + 1 (for the primary) but there is no hard rule on that. So what would 'sufficient' mean in this context? The chart with worker processes you showed does not provide much value and I am thinking of removing it as it's a time series that actually always has a discrete value - this is not volatile - so the bars don't make much sense. Please let me know if the prospect can provide a use case where the number of workers fluctuates.
Prospect says the sometimes he found that the number of worker are lees then he expected to be and lower than his threshold, may be a machine with less cpus than needed was allocated can cause that? any way the question was if there is a way to set threshold on that and get alert?