* Update the infludb broker to not need a server so that it'll work on non-data node servers
* Update messaging broker to keep track of connect urls for replicas
* Pull out magic numbers for running CQs in the broker
* Typo fixes in comments
* Update the process_continuous_queries endpoint to not have gzip
* Remove previous implementation of MeasurementNames in favor of new simpler one
* Fix run to work with CQ broker
* Fix CQ handler
* Fix SetTimeRange to use RFC3339Nano
* Fix the time range tests
* Fix the parser to parse for RFC3339Nano literals in addition to the other format
* Add logic for running CQs
* Remove duplicate WriteBufferSize default setting from config
* Update defaults on Broker for when to trigger CQ runs
* Add config settings for how often to calculate and recalculate CQ results
* Update Server to have CQ settings
* Update AST to fold removed time literals when setting time range
* Add periodic CQ running functionality to server
* Create influxdb specific broker that will periodically start continuous query calculation. Will send to the same data node as long as it's up and responding
* Add endpoint in handler to start processing of continuous queries on a data node
* Update SelectStatement to have GroupByInterval
* Update parser to verify that CQs have GroupByInterval if Aggregated
* Implement CreateContinousQuery on Server
* Add ContinuousQueries to database
It is quite possible that multiple applyCreateFieldsIfNotExist messages
will sit unapplied in the Raft commit log, each carrying exactly the same
information about new fields. So during application of this message only
change the Metastore if the fields actually changed.
Without this change, new shard group information is not persisted in the
metastore and only when the next database-save took place would this
update be persisted. If the server shutdown before the update, the new
shard group would be lost.
Paul Dix
Frontend
Philip O'Toole
Philip O'Toole
David Norton
Ben Johnson
Cory LaNou
Todd Persen