You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
The current design of the extension RPC server has it piggy backing off the general worker gRPC server. The lifetime mismatch between individual workers and this server causes issues during job host transitions. We will need explore one of two options:
Improve lifetime tracking with the existing RPC server
Only unregister endpoints when functions have finished draining.
Support multiple job hosts being registered. Will need to include an identifier for what set of endpoints an extension call needs to be routed to.
OR move these endpoints into a new server that lives in the job host.
Will need to measure the impact an extra server startup will have.
the-rule, AlexanderDerck, goldmelodyvn, Frulfump and CSymes