Sensors
Sensors wrap an Esper CEP engine with sharding logic to load EQL statements and consume PerfEvents that are produced by the collectors. If threshold/statements are violated, then OpsEvents are produced and opamp consumes to produce ActionOrders or WorkOrders.
The sharding logic does a mod of the manifest ID and poolsize global var to load statements from sensor_ksp on DAQ/Cassandra and consume from an opsmq perf-in-q-<shard>
.
re-shard
- Change the
SENSORPOOLSIZE
global var.
- Disable opamp (soon to be: disable sensor heartbeat monitoring).
- Commit and deploy only the sensor.
- While the sensor is bootstrapping (takes ~10 minutes), touch the update daq collector-artifact and daq logstash. (The env var is set in an attachment for logstash.)
- Commit and deploy (takes ~10 minutes).
- Verify that the queues on opsmq are clear and that the number of unhealthy components is normal.
- Enable opamp / sensor heartbeat monitoring.