second repeat :-) It works now. As you've said, it was a firewall problem, where the UDP ports below 1024 were blocked for all interfaces, so eth2 was also affected.
I've learnt today that ngrep reads the network data before it may be filtered by iptables. Best regards Peter Lars Ellenberg schrieb: > On Sun, Sep 27, 2009 at 07:07:16PM +0200, Peter P GMX wrote: > >> The strange thing is: We can see in the ngrep logs below that both >> machines receive communication on upd port 694 . >> That was the first I crosschecked, as this is not the first machine we >> setup sucessfully. >> And if the firewall lets through the messages on port 694, as we can see >> on both macines, then it should be fine, right? >> So my question is: Are there any other ports or protocols involved by >> this beyond udp port 694 and ICMPs? >> > > "firewall" may also be iptables on the cluster node themselves. > of course you will still see packets in tcpdump and other such tools. > > iptables-save | less > > >>> People tell me they don't have one. I repeat my advice. This repeats 2 >>> or 3 times and eventually they find the problematic firewall - and >>> > > first repeat ;) > > >>> either open port 694 on it, or shut the firewall off. Then the problem >>> goes away. I cannot count the number of times this has happened. >>> >>> The reason why heartbeat starts the IP on both machines is that each is >>> sure it's in control of the world. Without heartbeat communication - >>> that can happen - and it will. It's called a split brain condition. It's >>> a bad thing. >>> > > _______________________________________________ Linux-HA mailing list [email protected] http://lists.linux-ha.org/mailman/listinfo/linux-ha See also: http://linux-ha.org/ReportingProblems
