-
Type: Bug
-
Resolution: Won't Fix
-
Priority: Major - P3
-
None
-
Affects Version/s: 3.0.12, 3.2.8
-
Component/s: Sharding
-
None
-
ALL
-
Sharding 2016-08-29, Sharding 2016-09-19
The issue is because the explain command does not check that writes are to the config servers (which the real write command does). This will eventually result with the DBClientMultiCommand calling say() on SyncClusterConnection multiple times, which in turn calls say on the 3 connections inside SyncClusterConnection. This results with the connection calling say() multiple times in a row (w/o recv) which is not supported and the assertion ("MessagingPort::call() wrong id got:") is triggered when recv is called.
In v3.2, a different kind of assertion that happens much earlier, complaining that it is expecting a replica set name from the connection string (it fails because SCCC is not a replSet connection)
- related to
-
SERVER-27864 SlaveOK reads against SCCC config server fail with error "all servers down/unreachable when querying"
- Closed