NFS errors for fstab entries during boot time RHEL5
 

I have a problem in which about 4 of the 7-8 nfs mount entries in /etc/fstab do not get mounted during bootup.
I can execute `mount -a` via the command line an it works fine. When I watch the startup scripts execute I see "no route to host" error messages. This indicates that the routing tables have not be fully initialized before the nfs entries in the fstab file are attempted. I added a sleep command in the network startup script worked as well. Does anyone have a better solution to this startup issue?


Rick V

Well the routing tables should be fine, you've either got them up or you haven't, there's nothing to wait for in terms of pure time. You can try moving the netfs script later in the boot process maybe, but i'd be reckoning there's something specific causing this. I've seen this on a bunch of systems I had to take over where the users had just totally rearranged all the scripts, when in fact the issue was that they were using trunked network ports and the Cisco switch they were connected to was insisting on a full 50 second stp timeout before it'd forward packets. Unlikely that it's exactly that, but it did happen to me, and took me a long time to really suss out, but I fixed it properly on the switch side instead of fudging the init scripts. You can certainly add some temporary debug scripts to the equation, e.g. dumping the routing tables to a file within the start() of the netfs script.

During the initial trouble shooting I was trying to replicate the "no route to host" error message and after a series of unsuccessful ifconfig and service network commands I received the same message when I did a service network stop and tried the mount -a command ( I had previously umounted the nfs mounts in my fstab file). Before that I downed the interface, intentionally gave a unresolvable hostname for the mount, removed the host for the /etc/hosts file etc but received different messages. After I downed the network service the routing table via the netstat -nr command showed no route to the local subnet. Issuing the mount command at that point produce the same message. We have portfast enabled on all the switch ports which should prohibit the timing issue for STP forwarding. When I get more maintenance time on the node I will try to have the routing table dumped to output before the fstab entries are mounted.

I verified that portfast was enabled on the Cisco ports, I have inserted the retry=3 option into the fstab mount entries and everything gets mounted correctly now. There still must be some timing issue with the startup scripts that are causing these errors. We are going to leave the retry options into the fstab from now on to avoid any timing issues.