THIS IS A TEST INSTANCE. ALL YOUR CHANGES WILL BE LOST!!!!
...
Query Time range | Resolution of returned metrics | Comments |
---|---|---|
Last 1 hour | 1 min aggregates | Precision data available on demand as well |
Last 4 hours | 30 sec resolution for 1 hour of data, 1 min aggregates thereafter | |
Last 10 hours | 1 min aggregates | Honor query result limit |
Last 24 hours | 1 hour aggregates | min and seconds precision available if requested |
Last week | 1 hour aggregates | |
Last year | Daily aggregates | If query limit is reached query weekly aggregates |
Aggregation
- The granularity of aggregate data can be controlled by setting wake up interval for each of the aggregator threads.
- Presently we support 2 types of aggregators, HOST and APPLICATION with 3 time dimensions, per minute, per hour and per day.
- The HOST aggregates are just aggregates on precision data across the supported time dimensions.
- The APP aggregates are across appId. Note: We ignore instanceId or APP level aggregates. Same time dimensions apply for APP level aggregates.
- We also support HOST level metrics for APP, meaning you can expect a system metric example: "cpu_user" to be aggregated across datanodes, effectively calculating system metric for hosted apps.
- Each aggregator performs checkpointing by storing last successful time of completion in a file. If the checkpoint is too old, the aggregators will discard checkpoint and aggregate data for the configured interval, meaning data in between (now - interval) time.
- Refer to Phoenix table schema for details of tables and records.