Fix major compaction metrics to not reference the tserver

[dlmarion]

Closes #5062

[keith-turner]

Seems like the three metric names in this PR could use the accumulo.compaction prefix like in #5075. Like accumulo.compaction.majc.queued. Also not something for this PR, but it seems like MAJC_QUEUED and MAJC_RUNNING could be removed. Those two metrics are dupicating data the is present in per compactor RG metrics and could be derived from those.

[dlmarion]

Seems like the three metric names in this PR could use the accumulo.compaction prefix like in #5075. Like accumulo.compaction.majc.queued.

I merged into #5075 and fixed the names here.

Also not something for this PR, but it seems like MAJC_QUEUED and MAJC_RUNNING could be removed. Those two metrics are dupicating data the is present in per compactor RG metrics and could be derived from those.

CompactionCoordinator.registerMetrics registers MAJC_QUEUED and MAJC_RUNNING. I see that QueueMetrics registers COMPACTOR_JOB_PRIORITY_QUEUE_JOBS_QUEUED with a queue.id tag, which could be summed up to replace MAJC_QUEUED. I don't see a direct replacement for MAJC_RUNNING though. What did you have in mind for that?

[dlmarion]

CompactionCoordinator.registerMetrics registers MAJC_QUEUED and MAJC_RUNNING. I see that QueueMetrics registers COMPACTOR_JOB_PRIORITY_QUEUE_JOBS_QUEUED with a queue.id tag, which could be summed up to replace MAJC_QUEUED. I don't see a direct replacement for MAJC_RUNNING though. What did you have in mind for that?

Ok, I see now in #5062 you said ... this information can be computed by summing idle metrics from compactors. I'm wondering if that may not be intuitive to a user to use that for running compactions given the name.

[dlmarion]

CompactionCoordinator.registerMetrics registers MAJC_QUEUED and MAJC_RUNNING. I see that QueueMetrics registers COMPACTOR_JOB_PRIORITY_QUEUE_JOBS_QUEUED with a queue.id tag, which could be summed up to replace MAJC_QUEUED. I don't see a direct replacement for MAJC_RUNNING though. What did you have in mind for that?

Ok, I see now in #5062 you said ... this information can be computed by summing idle metrics from compactors. I'm wondering if that may not be intuitive to a user to use that for running compactions given the name.

Ok, using the server idle metric to gauge whether or not the Compactor is running a compaction is probably not the way to go. The process has to be idle for so long, regulated by Property.GENERAL_IDLE_PROCESS_INTERVAL, to be considered idle. This is so that the metric is not constantly flapping.

[dlmarion]

I removed MAJC_QUEUED and MAJC_RUNNING. I also added some new metrics as at least one was being used in the wrong place. I look at this as some general cleanup, but I think more might be needed. Specifically, I'm thinking the metric names might need to be reviewed and corrected.