Hide Forgot
Description of problem: When trying to automount a NFS directory on a very busy server, I sometimes get: May 27 05:21:06 cynosure automount[4026]: attempting to mount entry /data/sw1 May 27 05:21:48 cynosure automount[24502]: >> mount: RPC: Timed out May 27 05:21:48 cynosure automount[24502]: mount(nfs): nfs: mount failure alexandria:/export/data1 on /data/sw1 May 27 05:21:48 cynosure automount[24502]: failed to mount /data/sw1 May 27 05:21:48 cynosure automount[24502]: umount_multi: no mounts found under /data/sw1 I would at least like the option of having this wait longer (forever?). Perhaps this can be changed with an RPC parameter (for longer/no timeout), or just more retries. Version-Release number of selected component (if applicable): autofs-4.1.4 How reproducible: fairly Steps to Reproduce: 1. try to automount directory from very busy nfs server Actual results: mount fails. Accessing program gets a no such file error. Expected results: mount keeps trying until it succeeds. Program eventually gets to access the file and everyone is happy.
I'll look into this. My current thinking on the matter is that there will be a global configuration setting in /etc/sysconfig/autofs that will determine the behaviour for all requests. If you need finer grained control of this, please let me know as it will affect the design.
At this point I think it could be global without trouble. Might be nice for per automounted directory if that wouldn't be hard, but I suspect that if you want mount to always succeed, you want mounts to always succeed everywhere.
Upon reconsidering this, why don't you simply specify a retry count on your mount command line? man 2 mount: Mount options for nfs retry=n
man 5 nfs: retry=n The number of minutes to retry an NFS mount operation in the foreground or background before giving up. The default value is 10000 minutes, which is roughly one week. The default is 1000, so this doesn't appear to apply here.
This really should be an NFS setting. If you want the default timeout to be higher, file the bug against nfs-utils. Perhaps the advertised default isn't implemented correctly? As it stands, autofs uses the defaults. You can override these defaults in your maps if you so choose. However, I stand firm in my conviction that this is not an automount problem.