This is a simple network diagnostic problem.
Start by determining how far the packets are getting. Use tcpdump on the sending machine to make sure the packets are getting generated and sent, then tcpdump on the receiving machine to see if the packets are arriving. If there are intermediate routers or firewalls, you can tcpdump from those too (all decent appliance routers have some sort of capability for instrumenting at least flows, if not individual packets). With a managed switch, you can use a mirror port to see if the packets are going through the switch (although for a simple problem like this, it should be unnecessary). Worst case, if you suspect an intermediate hop, a traceroute can be a reasonable diagnostic tool, and if it's a TCP service that's causing problems, then a tcptraceroute is about 100 times better.
Once you've determined how far the packets are getting, you can examine the offending machine itself. The firewall at one end is by far the most likely candidate;
iptables -L -v will show packet counts alongside the rules, to show which rules are getting used. If a firewall rule is actually getting hit, it should show up fairly clearly. If you have default DROP rules (a good idea), you should also put a LOG message in the chain to show when the default is being hit.
Given that you haven't provided any description of what you're actually trying to do (ping, connecting to a TCP service, whatever), nor what exactly constitutes "not accessible", it's tough to describe what else to check. For a TCP service, ensure that there's no tcpwrappers getting in the way, and that the service is bound on the correct port and IP (
Beyond that, more specific diagnostic details will require more details of the problem. Be very specific as to what you're trying to do, and maybe someone can provide you the right answer.