I am running JBoss 3.2.4 in the all configuration in a cluster of two nodes using auto discovery. The services binds to a specific IP (started up using -b). Eveything works fine for the first few minutes, then suddently some HA-JNDI lookups on IP:1100 starts to take 60 seconds, some even 1088 seconds to complete.
Eventually the correct value is retrieved from the HA-JNDI tree, but it takes random times. Some calls return immediately, some takes very long. The two servers are on a 1Gb network link.
The code regularly binds/unbinds/reads/removes data from the tree - I am using it for locking of entities cluster wide.
Why would this happen? I know the call to initialContect.xxx() such as initialContext.unbind() takes very long.
Use JBossCache instead. JNDI was not designed for high write-read ratios (most read-only)