Question about cluster mode

[yharish991]

Does this library monitor all the processes in the cluster mode? we are running in cluster mode with let say 8 node instances, will it send metrics for each node process, or only from master process?

[mattcolegate]

Hi @yharish991, appmetrics will only monitor individual node instances, so each of your 8 would have to have separate appmetrics-dash instances. I believe that Bluemix Availability Monitoring works with clusters, have a look at https://console.bluemix.net/catalog/services/availability-monitoring Hope that helps!

[sam-github]

@mattcolegate the OP is using node cluster. I believe the -dash will not work under this configuration, because a single TCP/HTTP port is going to randomly receive requests from different appmetrics instances. I don't think this can be resolved easily, without using a different architecture, one in which the dashboard isn't served by an application instance.

I'll confirm it doesn't work when I have some time.

[sam-github]

In the meantime, what are the other options for subscribing to appmetrics data: I think appmetrics-elk and/or appmetrics-statsd would be better, with an external dashboard, it would be getting information from each appmetrics instance independently, and be able to display the data and the process it originates from.

[sam-github]

also, IBM APM and IBM BAM (application monitoring) are full featured systems tracing and monitoring solutions

[gustavomassa]

I'm facing the same issue using cluster mode, I think the best idea is to make app-dash able to receive multiple metrics package based on an id/name, so each worker can send its own data. We could have a combobox or a list on the dashboard to select which worker we want to see metrics data...
This idea is also applicable for multiple services sending metrics data to one monitoring service, so we can have a combobox list with the service name, and just select which one to show the graphs..

We could start just adding an id/name identifier to the data packages sent to front-end.

[tobespc]

having the dash be able to split between different workers is an interesting idea , and not a direction we had previously considered. We do provide integration hooks to prometheus so that would also give you the view from each worker node if you wanted something that would work right now