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
In the course of addressing #12823, I was faced with the error message No reply handler for reply with id: 116. This means that (typically) the back end sent a reply for a request that the current instance of RPCProtocol does not know anything about. Unfortunately, this error message is very unhelpful, because it's impossible to find out what request caused this and whether the request was sent by the same RPCProtocol that the reply got routed to.
My proposal would be to introduce optional tracing messages to our RPC protocol that include:
A globally unique message id (currently, the message id is per RPCProtocol instance
The ID of the sending RPCProtocol instance (or other identifying info about the originator, like service id)
The request that cause the reply. This info could contain a configurable amount of information like
the request method
the request parameters
the stack trace of when the request was sent
Since there will be a performance penalty for this tracing, we should discuss how exactly this tracing could be configured: do we need to be able to turn this on "on the fly" or can it be a cli parameter or even a compile time config?
The text was updated successfully, but these errors were encountered:
Feature Description:
In the course of addressing #12823, I was faced with the error message
No reply handler for reply with id: 116
. This means that (typically) the back end sent a reply for a request that the current instance ofRPCProtocol
does not know anything about. Unfortunately, this error message is very unhelpful, because it's impossible to find out what request caused this and whether the request was sent by the sameRPCProtocol
that the reply got routed to.My proposal would be to introduce optional tracing messages to our RPC protocol that include:
RPCProtocol
instanceSince there will be a performance penalty for this tracing, we should discuss how exactly this tracing could be configured: do we need to be able to turn this on "on the fly" or can it be a cli parameter or even a compile time config?
The text was updated successfully, but these errors were encountered: