HA during war deployment and 404

aandrius Jan 2, 2015 11:02 AM

Hello,

our setup consists of 4 wildfly 8.1 instances managed via domain, with nginx dispatching incoming requests to wildfly backends. Everything works fine, unless we try to deploy .war file.

Now we see that once we deploy .war file, wildfly instances report HTTP 404 error and nginx sees these backends as unavailable during deployment time. Since we want to achieve HA even during deployment we expect only one wildfly instance/backend being down during deployment. Also we found no way to setup custom error code than default 404 to report from wildfly during deployment, therefore we cannot setup logic "try proxy_next_upstream if http_503"

Questions:

1. How to make Wildfly report other than 404 error code during deployment, say 503 ?

2. how to achieve HA even during deployment so that only one instance/backend is disabled ?

Btw, is there anything like /application-status or /cluster-status built-in into wildfly so that we can see what's going on with our cluster AND/OR application?

Thank you very much,

-Andrius..

1. Re: HA during war deployment and 404

pferraro Jan 7, 2015 11:47 AM (in response to aandrius)

This is a common shortcoming of most load balancers - they have no concept of web applications within a node. If a node is deemed to be active, it will route requests for any application to that node, completely ignorant of the status of that web application on the node.
This specific issue was one of the primary drivers for the development of mod_cluster. mod_cluster tracks the state of each application, rather than assuming that all applications on a server are in the same state. Thus mod_cluster will never direct a request for an application to a given node unless that target application is deployed on that node.
Actions