Sunday 15 February 2015

jboss - Prevent deployment to entry node, only deploy to other nodes -



jboss - Prevent deployment to entry node, only deploy to other nodes -

i have free openshift business relationship default 3 gears. on have installed wildfly 8.1 image using openshift web console. set minimal , maximal scaling 3.

what happens openshift create 3 jboss wildfly instances:

one on entry node (which running haproxy) one on auxiliary node one on auxiliary node

the weird thing jboss wildfly instance on entry node default disabled in load balancer config (haproxy.conf). but, openshift still deploying war archive whenever commit in associated git repo.

what's problematic here because of incredibly low number of max user processes (250 via ulimit -u), jboss wildfly instance on entry node cannot startup. during startup jboss wildfly throw random 'java.lang.outofmemoryerror: unable create new native thread' (and no, memory fine, it's os process limit).

as result, deployment process hang.

so summarize:

a jboss wildfly instance created on entry node, disabled in load balancer jboss wildfly in default configuration cannot startup on entry node, not trivial war. the deployer process attempts deploy jboss wildfly on entry node, despite beingness disabled in load balancer

now question:

how can modify deployer process (including gear start command) not effort deploy jboss wildfly instance on entry node?

when app scales 2 gears 3, haproxy stops routing traffic application on headgear , routes 2 other gears. assures haproxy getting cpu possible application on headgear (where haproxy running) no longer serving requests.

the out of memory message you're seeing might not actual out of memory issue bug relating ulimit https://bugzilla.redhat.com/show_bug.cgi?id=1090092.

jboss openshift wildfly-8

No comments:

Post a Comment