We run Linux (Ubuntu 12.04) on the 10 or so desktops at my workplace. After installing a new machine today, I found an odd problem. It could not communicate (even ping) with some other boxes on our LAN. All of our boxes are in the same subnet (192.168.0.0/24). All of the other boxes can ping each other fine, it's just this new one that can't ping some of the boxes. It can ping some fine with 0% packet loss, others with 50 to 90% packet loss and some with total packet loss. Interestingly, the other hosts seem to be able to ping the new box and get an echo back even when it can ping them (can't ping from all of the hosts because some of them are network devices like printers). One of the hosts that it struggles to ping is the router/gateway. It can ping our main server (also does DHCP and DNS) with no problems. I haven't tested pinging the outside world (WAN) much, but no serious problems there. I doubt it's a problem with the volume of data because I'm able to get NFS working on there and push a lot of traffic without issue.
The first thing we do when installing a new desktop is applying our standard puppet manifest to it, so we can be sure that they all have a near identical configuration.
We only have one switch, and that's connected to only one router and everything is connected to that switch. I'm pretty satisfied that it's not a physical problem because the new box can ping some hosts find and the switch is in between the box and all of the other hosts. All wiring is copper Cat5e cable and I have tested the connection with another box to make sure it's not the connection.
I have seen this once before on similar hardware and it was fixed by a reinstall, so I can try that if all else fails, but I'd like to know what's actually wrong.
I don't even know where to start when it comes to diagnosing this. Any suggestions?
[link] [3 comments]