We have an elastic cluster with a huge watcher count. Because of that, we have a continuous search load with peaks, but able to handle it. In the perspective of a growing number of watchers, we are not sure if the current implementation is really scaleable.
I can´t find any source explaining the watcher workflow and looking for more in detail knowledge. The questions are the following:
- Who manages/orchestrates the execution of a watcher? Is it the current master node?
- Which node issues the watcher execution? Again the current master?
- On which node the search, transform and action phases are executed? Master?
- Is there any option to configure the watcher management/execution node?
- Is there any limitation in concurrent watcher execution? For example, if at a given time there are many watchers planned to execute, what is the order of execution, and is the workers/thread count limited?
References/Sources would be very appreciated, thank you in advance!
