Details
-
Improvement
-
Resolution: Fixed
-
Major
-
None
-
None
-
None
-
1
Description
When we detect an EOF on a connection when we go through all outstanding requests and call their callbacks with EOF (unless an upstream component decides that it should be a different error).
We should change this so that the DCP queue is drained first (which means that any outstanding End messages will be handled and an EOF not returned), but only when the connection is closing due to external initiation.
Attachments
For Gerrit Dashboard: GOCBC-1242 | ||||||
---|---|---|---|---|---|---|
# | Subject | Branch | Project | Status | CR | V |
170813,3 | GOCBC-1242: Flush the DCP buffer on EOF originating from remote | v10 | gocbcore | Status: MERGED | +2 | +1 |
170880,3 | GOCBC-1242: Flush the DCP buffer on EOF originating from remote | master | gocbcore | Status: MERGED | +2 | +1 |
171475,2 | GOCBC-1248: Check if client already closed during graceful shutdown | master | gocbcore | Status: MERGED | +2 | +1 |
171513,2 | GOCBC-1248: Check if client already closed during graceful shutdown | v10 | gocbcore | Status: MERGED | +2 | +1 |