Marklogic socket connection error

Question

Hi Markloggers out there,

we have a server setup that produces the following error log lines which we do not understand nor find any usefull documentation online:

2014-09-12 14:41:35.828 Error: PooledThread::run: SVC-SOCCONN: Socket connect error: wait 185.10.49.71:34932: Timeout
2014-09-12 14:41:45.847 Error: PooledThread::run: SVC-SOCCONN: Socket connect error: wait 185.10.49.71:34933: Timeout
2014-09-12 14:42:05.891 Error: PooledThread::run: SVC-SOCCONN: Socket connect error: wait 185.10.49.71:34935: Timeout
2014-09-12 14:42:16.900 Error: PooledThread::run: SVC-SOCCONN: Socket connect error: wait 185.10.49.71:34937: Timeout
2014-09-12 14:42:27.915 Error: PooledThread::run: SVC-SOCCONN: Socket connect error: wait 185.10.49.71:34939: Timeout
2014-09-12 14:42:38.935 Error: PooledThread::run: SVC-SOCCONN: Socket connect error: wait 185.10.49.71:34941: Timeout
2014-09-12 14:42:58.983 Error: PooledThread::run: SVC-SOCCONN: Socket connect error: wait 185.10.49.71:34945: Timeout
2014-09-12 14:43:09.996 Error: PooledThread::run: SVC-SOCCONN: Socket connect error: wait 185.10.49.71:34947: Timeout
2014-09-12 14:43:21.008 Error: PooledThread::run: SVC-SOCCONN: Socket connect error: wait 185.10.49.71:34949: Timeout
2014-09-12 14:43:32.027 Error: PooledThread::run: SVC-SOCCONN: Socket connect error: wait 185.10.49.71:34951: Timeout
2014-09-12 14:43:43.033 Error: PooledThread::run: SVC-SOCCONN: Socket connect error: wait 185.10.49.71:34953: Timeout

Anybody a pointer?

thx

Is this part of a cluster? Is 185.10.49.71 the server you're running on, or something that this server is trying to use? — Dave Cassel

mblakele mblakele · Accepted Answer · 2014-09-12T15:07:23

Those are low-level socket connection errors. There's a low-level network communication problem. It could be just about anything: you'll need to gather more information.

I'd start by checking the OS logs. There may be something there. Try to find out what should be talking to that IP address and why. Is it a call to xdmp:http-get? Or MarkLogic intra-cluster communications? Or is it just a client request that was dropped or timed out?

If this is a cluster, you may have a problem where an OS firewall or network firewall is terminating long-running connections. If so you'll have to find and fix that. Ideally there shouldn't be any network-layer firewall or other potential sources of latency in between the hosts in a cluster. If a host-level firewall seems necessary, you can whitelist the cluster IP addresses and ports 7997-7999.

Marklogic socket connection error

1 Answers