Red Hat Bugzilla – Bug 441463
netfs starts too early after networking
Last modified: 2013-07-18 16:59:27 EDT
Description of problem: On boot. Netfs starts too early after networking, this causes nfs shares to fail because the host is not resolvable while networking is starting. Version-Release number of selected component (if applicable): initscripts-8.45.17.EL-1.el5.centos.1 How reproducible: boot server with an nfs share. Actual results: on boot - when start scripts are scrolling it says. Mouting NFS filesystems mount : mount to NFS server "192.168.0.102" failed: system error : no route to host once server fully booted. Log on and then type "mount -a" the NFs share mounts fine then. Expected results: It should mount it fine Additional info: The fix : edit /etc/init.d/netfs Add this line : action $"Sleeping for 60 secs: " sleep 60 So the file now reads : [ ! -f /var/lock/subsys/portmap ] && service portmap start action $"Sleeping for 60 secs: " sleep 60 action $"Mounting NFS filesystems: " mount -a -t nfs,nfs4 Website : http://crazytoon.com/2007/05/11/centos-and-redhat-problem-nfs-mount-at-boot-up-fails-with-error-system-error-no-route-to-host/
Moving NTP will fix the problem. I hadn't realised there was already an option to synchronise time before starting ntpd. With that option turned on, as long as ntpd starts before netfs, this bug goes away.
This ALSO SHOWS up on a system that is an NFS client of itself, because the NFS server does not startup until way later after netfs ties to mount the NFS moint points in /etc/fstab. NTP will *NOT FIX* this issue. Netfs needs to come after the nfs server startup.
My comment appeared on the wrong bug. Must have made a booboo sorry, My problem was specific to ntfs4, and this comment is on the wrong bug - sorry.
This is still an issue in RHEL 5.5. Updated a system to rhel 5.5 and had to manually reorder the init scripts from the default for a nfs server that mounted shares from itself.
While I encountered some old machines, the real issue appears to be some timeout waiting for the arp table to properly fill. So I added the following lines below portmap start: # bug 441463 ping -c 1 192.168.102.1 >/dev/null action $"Waiting for gateway to appear: " eval "arp 192.168.102.1 >/dev/null"
If NetworkManager is enabled, does setting NETWORKWAIT=yes in /etc/sysconfig/network fix the issue? That will wait up to 30 seconds for a network connection to start, and will exit when the shorter of (a) 30 seconds or (b) a network connection has occurred. That should ensure that the NFS server can be found before any further scripts can proceed.
Has anyone been able to check whether the NETWORKWAIT=yes fixes the issue?
Due to lack of response for 5 months closing the issue for now. If this is super critical and has to be fixed for the 5.10 or 5.11 release then please reopen it with enough information to qualify this issue as a critical item for the release. Thanks