There is also a bug in RHEL4U2 and earlier where using autofs you could hit a race condition that actually fails a mount.

If at the same time an autofs mount times out and you go to use a file on the same mount, then it looks like the mount failed and you get some type of file access error. If you check immediately after that the file is fine.

When you are doing grid-related jobs that take hours-days-weeks, it really stinks to hit this bug.

It supposedly only affects "ghosted" auto-mounts. A workaround I use is to use --timeout=0. It half-defeats the purpose of autofs, but when you have lots of mounts, it sure makes it easy.

--tmac

Graeme Fowler wrote:

Hi

On 05/07/2006 09:11, Josef Radinger wrote:

...

We have a bunch of linux-systems, which get some mounts from two netapp-filers FAS3020. both filers run under ontap 7.0.3

<snip>

http://www.netapp.com/library/tr/3183.pdf has several useful recommendations, particularly in section 4.

I've had great success (and no hangs) using FC1, FC2 and FC4 against NetApp filers (although I have no access to those systems at present!) and if I recall correctly, I used a combination of mount options something like:

filer:/vol/vol1/export /export nfs rw,vers=3,hard,intr,tcp,timeo=600,retrans=2,rsize=32768,wsize=32768,bg

That gave me no problems whatsoever.

The document recommends very strongly that you use TCP as your transport protocol instead. Have a read and a play, and see what you find.

Graeme