[Mageia-dev] Any progress on the NFS mount problem? - One step further

Mon Oct 1 19:45:37 CEST 2012

Le 01/10/2012 17:26, Anne Wilson a écrit :
+> -----BEGIN PGP SIGNED MESSAGE-----
+> Hash: SHA1
+>
+> On Monday 01 Oct 2012 13:18:36 Guillaume Rousse wrote:
+>> Le 22/09/2012 21:16, Anne Wilson a écrit :
+>>> udp        0      0 *:nfs                       *:*
+>>>
+>>> unfsd
+>>
+>> Ouch... You're running an obsolete alternative nfs server. You
+>> should uninstall the unfs3 package, and this package should get
+>> removed
+> from
+>> the distribution.
+>>
+>> To summarize, here are the mandatory steps: - check you have
+>> nfs-utils-clients installed on clients - check you have
+>> nfs-utils-clients and nfs-utils installed on server - check
+>> rpcbind, and nfs-common services are running on clients - check
+>> rpcbind, nfs-common and nfs-server services are running on
+> the
+>> server
+>>
+> Not sure of the command-line for checking things that don't respond to
+> "service whatever status", but MCC says all those services are running
+> on the server.
+MCC won't give any useful detail, "service nfs-common status" will...
+
+[..]
+>> Now, to ensure usage of NFSv4 instead of obsolete NFSv3: - check
+>> every host is configured to run rpc.idmapd, meaning NEED_IDMAPD=yes
+>> in /etc/sysconfig/nfs-common
+>
+> Done
+>
+>> - check your server export a root filesystem, using fsid=0 options
+>> in its /etc/exports configuration file
+>
+> Added that - how do I tell it to ignore the old exports and act on
+> those in /etc/exports?  I can't find my notes about that.
+Quick answer: service nfs-server restart
+
+>> - drop all your legacy nfs options, such as wsize and rsize from
+>> your mount options
+>>
+> Done - the lines now match the working QNAS mounts, eg
+>
+> 192.168.0.40:/home /mnt/borg2/home nfs user,timeo-14 0 0
+Why do you need a timeout, and why 14s ? If you're not an NFS expert, 
+you'd better rely on default values only.
+
+Also, take cares, with NFSv4, you don't mount the actual server 
+filepath, but the exported filepath: the root filesystem (the one with 
+fsid=0) will be available as /, whatever its actual path.
+
+Once again, test everything manually with manual mount command before 
+hardcoding them in /etc/fstab.
+
+>> Last, you'd better use autofs to mount those filesystems on
+>> demand, rather than hardcoding them in /etc/fstab, which avoid to
+>> relies on server availability during the boot.
+>
+> Replace nfs with autofs in those lines?
+No, that means dropping any nfs filesystem entry in your /etc/fstab, and 
+configuring autofs daemon instead. But that's off-topic right now...
+
+-- 
+BOFH excuse #260:
+
+We're upgrading /dev/null
+