Mongos abrupt shutdown - Very frequent, Mongo DB version 3.2.4

[pavan....@outlook.com]

We are running  3Shard Mongo Cluster in our lab - QA environment before we go live.  

The configuration is simple and straightforward.  

3 shard system  -  each shard with its replica systems - 3 servers, 2 backing each other and the 3rd one as arbiter. 

The config server is running in a non replica set mode, given the setup is experimental. 

Mongos is running on its own instance. 

All servers are in AWS 

--------------------------------------------------------------------------------

3 - Shard System : M3.Medium - 4 Core/15 Gig 

failover and Arbiter on a lower configuration 

----------------------------------------------------------------------------

Config Server on a M3.Medium - 4 Core/15Gig

---------------------------------------------------------------------------------

Mongos Server on a t2.Medium  - diskless 

--------------------------------------------------------------------------------------------

Mongos seems to be shutting down with no details or errors .  All we see in the log is 

 Can't use connection pool during shutdown

2016-05-03T04:43:02.586+0000 I SHARDING [signalProcessingThread] dbexit:  rc:0 

1) What is not clear from this message is, is the app trying to get a connection and Mongos can't serve it ?  

2) Is Mongo unable to get connections from Config Server ?  -- No indication of errors or whatsoever in the logs .. 

3)  Is Mongos failing to get connections to Shard Servers ? ---  There is no such indication in any of the shard logs .. 

Has anybody faced similar situation with Mongo Cluster ?  We are going to move the mongos to a multiple servers for redundancy, most likely deploy Mongos on multiple app servers and load balance it.   We would however like to know what may be causing the shutdown. 

Thanks

[Rhys Campbell]

Check your /var/log/messages or equivalent. Perhaps the OOM killer is being invoked.