[EnMasse] Disconnect from time to time

Gordon Sim gsim at redhat.com
Wed Mar 27 10:46:08 UTC 2019


On 27/03/2019 10:23 am, Ulf Lilleengen wrote:
> According to the admin pod logs it looks like there seems to be only 1. 
> Also this is a sharded topic broker from looking at the broker logs.
> 
> Ignoring all the tracing stuff, i can't find any notion of disconnection 
> happening from the router side in the logs. From the broker log:
> 
> 2019-03-27T03:40:41.970Z INFO  [ProtonClientConnectionManager] 
> Connection messaging-aloxy/172.30.9.50:56671 <http://172.30.9.50:56671> 
> destroyed
> 
> Whereas the 2 entries in the router in the same time is:
> 
> 2019-03-27 03:40:37.417193 +0000 SERVER (info) [94473]: Accepted 
> connection to 0.0.0.0:55671 <http://0.0.0.0:55671> from 
> 10.128.0.67:36180 <http://10.128.0.67:36180>
> 2019-03-27 03:40:45.478579 +0000 SERVER (info) [94474]: Accepted 
> connection to 0.0.0.0:56711 <http://0.0.0.0:56711> from 10.128.0.1:36070 
> <http://10.128.0.1:36070>
> 
> There are these messages from the lost link I guess:
> 
> [0x7f7af4085850]:1 <- @detach(22) [handle=8, closed=true, 
> error=@error(29) [condition=:"qd:routed-link-lost", 
> description="Connectivity to the peer container was lost"]]

The connection above is an inter-router connection. (Topics don't use 
co-located routers any more, do they?). It is that other router that 
seems to be initiating the detach. I assumed there were two routers and 
the broker had disconnected from the other one.

Running qdstat -n against the router would show any other routers.

> [0x7f7ae8190e20]:0 -> @detach(22) [handle=0, closed=true, 
> error=@error(29) [condition=:"qd:routed-link-lost", 
> description="Connectivity to the peer container was lost"]]
> 
> I wonder if the improved logging in router 1.6.0 (in the upcoming 0.27.0 
> release) could shed some more light on this.
> 
> Best regards,
> 
> Ulf




More information about the enmasse mailing list