blocking problem "no physical address" with jgroups

[francesco.sclano]

Hi everybody,

I'm using:

1. infinispan 6.0.0.Final in REPL_SYNC mode

2. jgroups 3.4.1.Final configured with a TCP stack (see attached configuration)

3. java 1.7.0_45

4. the topology of my grid is composed by four nodes: A, B, C, D

    4.a) only node A write on cache in transactional mode (locking mode: OPTIMISTIC, isolation level: REPEATABLE READ - see attached configuration)

    4.b) C and D are behind a load balancer, this is the reason for using a TCP stack instead of an UDP stack

    4.c) all cluster nodes are listed in initial_hosts

Here the problem:

after a hot deploy I'm repeatedly getting following WARN message for 180 seconds:

17:16:30,325 WARN [org.jgroups.protocols.TCP] (Timer-5,ISPN,LXRD-WWIC-FE02-7602) JGRP000032: LXRD-WWIC-FE02-7602: no physical address for e4b906d6-f890-7118-2d64-be1460075b4f, dropping message

once these 180 seconds have passed, I got a different WARN message:

17:16:30,466 WARN [org.jgroups.protocols.TCP] (TransferQueueBundler,ISPN,LXRD-WWIC-FE02-7602) JGRP000032: LXRD-WWIC-FE02-7602: no physical address for e4b906d6-f890-7118-2d64-be1460075b4f, dropping message

finally my application doesn't log anymore and remains blocked, even above logs are WARN messages!

I attach both my infinispan and jgroups configuration.

Many Thanks

Best Regards

Francesco Sclano

[sannegrinovero]

Hi Francesco,

last time I saw such an error message, it turned out the network hardware was defective; on a similar occasion, it was the ethernet stack driver to be buggy. Do you have a chance to try your same configuration and stack on a different machine?

[francesco.sclano]

Hi Sanne,

I would like to thank you. Yes, the problem was as you supposed. Now it works.

[sannegrinovero]

Great! good to see this confirmed, as I've only been able to see it three times (counting you) in 5 years.