| Large HA PuppetDB may take up to an hour (maybe longer) to start due to the initial sync. During this time the HTTP metrics are not available and it is difficult to determine what the sync is doing and if it is progressing as expected. These are metrics that I use personally and are probably some of the more useful are (Bear in mind this is what infludb is calling them, won't map exactly):
- global_processed (To work out commands per second)
- global_processing-time
- queue_depth
- jvm-metrics_heap-memory_committed
- jvm-metrics_heap-memory_used
- jvm-metrics_cpu-usage
- global_message-persistence-time
- PDBReadPool_pool_Usage
- PDBReadPool_pool_Wait
- PDBWritePool_pool_Usage
- PDBWritePool_pool_Wait
Having these working as soon as the process starts instead of once it gets to a "running" state would be great, ideally none of the metrics would depend on the fact that PuppetDB is "running" as opposed to "starting" |
