Description
Make RELOAD and MAINTENANCE commands block until their tasks are complete. This gives feedback to the caller that it's now safe to execute next step in their workflow.
For RELOAD:
- Wait until active clients dropped the reference to the old pool. idle clients will hold it until the next query. Maybe add a notify to all clients to fetch latest
Cluster config immediately to make this more deterministic?
For MAINTENANCE:
- Wait until all clients are either
idle or waiting.
Description
Make
RELOADandMAINTENANCEcommands block until their tasks are complete. This gives feedback to the caller that it's now safe to execute next step in their workflow.For
RELOAD:Clusterconfig immediately to make this more deterministic?For
MAINTENANCE:idleorwaiting.