[daip] Re: AIPS tape locking query

[Patrick P. Murphy]

On Tue, 12 Sep 2000 14:56:29 +0100 (BST), Robert Laing
   <rlaing at astro.ox.ac.uk> said: 

> Our AIPS installation in Oxford (4 Linux PCs running 31DEC99) has been
> plagued for some time with an intermittent problem involving tape
> locking files. The symptom is that one execution of an AIPS task (e.g.
> FITTP or FITLD) will terminate correctly, but that subsequent tape
> operations will hang, with the message ZTPOPN: STILL WAITING FOR FILE =
> DA00:TPD001001;

Is the DA00 area in this case on a NFS-mounted filesystem?  If so, does it
have to be?  I've taken to making the $AIPS_ROOT/DA00/$HOST/ directories
into symlinks pointing at some area local to the host in question.  This
does an end run around any NFS-related file locking problems.

There should not be any such problems; the file locking in Linux for both
local and NFS mounted filesystems is now solid (thanks in no small part to
Jeff Uphoff's work on statd).  This applies to Red Hat 6.2; if you've got
an older or different release, you should check to see that you have
kernel NFS file locking enabled (and that lockd and rpc.statd are running
on the Linux system).  Nonetheless, I've seen some problems on an
intermittent basis; anytime I try to isolate the problem in a smaller
program, the darn thing works perfectly :-(  But in AIPS, trying a MOUNT
when the DA00 area is NFS-mounted from a Solaris system can often result
in the situation where the TPD file is locked and you have to change its
inode. 

Hope this helps.

				- Pat