Expose Server components over JMX

[shs96c]

We have the ManagedService annotation that used to be used expose the old selenium server via JMX. We should make the new server equally manageable.

[samueltauil]

I have interest on that to capture info from JMX and send to Prometheus then been able to build a custom chart

[amiluslu]

Can you give me a little bit details about this issue ? If you give me a little bit details or clue, hints, I can start to fix.

[shs96c]

@amiluslu we discussed this a little on the Slack channel. Do you still need some help getting started? I'm happy to provide more information if needed.

[pujagani]

The older Grid had the following managed services and attributes.

• Hub as a managed service.
  • Managed attribute :
    • Configuration
    • Session request count
• DefaultRemoteProxy as managed service :
  • Managed attributes:
    • Proxy status (alive or no)
    • It extends the BaseRemoteProxy. BaseRemoteProxy has managed attributes:
    • Max number of sessions
    • Remote host
    • Total number of used slot
    • Total number of slots
    • Proxy status
    • Resource usage counted as (total no of used slots/ max number of sessions) * 100
    • The last session start time
    • SeleniumServer which is an implementation of GridNodeServer as a managed service.

The new Grid can have analogous managed services and attributes as follows:

• Each server of grid servers can be a managed service :
  • DistributorServer
  • Hub server
  • SessionMapServer
  • Standalone server
  • NodeServer
  • RouterServer
  • NewSessionQueuer Server
  • EventBus
• All above-listed components share config values in BaseServerOptions. These config values can be exposed as managed attributes.
• Each individual components have additional component-specific attributes, which can be exposed as managed attributes.
• The Distributor is aware of the Grid model. Grid specific attributes can be exposed by leveraging Grid model.
• Information specific to a Node like max session count, slot information and availability can be obtained from via node implementations.

These are the observations and suggestions so so far. Can I get some help in understanding if this aligns with the intended feature? Please provide any other details I might have missed.

[diemol]

Fixed via #8838