[Info-vax] DNFS1ACP using 100% of CPU

[Sum1]

On 2013-10-13 02:21:09 +0000, JF Mezei said:

> <snip>
> 
> With VMS NFS behaving, are you able to access the web directories from
> DCL without triggering NSG going 100% CPU ? Can you do a dir/full on all
> the files the web server is likely to access ?

NFS has been running happily now for many years, after a 
well-documented set of configuration issues (in this group and on 
Hoff's site).  Al expected RMS functionality works fine :)
> 
> Does the OS-X directory contain VMS file semantic files ( for instance
> temp.txt would be accompanied by some other file (name escapes me now)
> which would contain the RMS attributes, and the VMS NSF client would
> then provide the VMS app with a temp.txt having  RMS semantics instead
> of a raw stream of bytes.

The OSX folders containing the static files have no semantic files nor 
do they have any hidden files.
> 
> There are settings when you mount a remote disk to decide whether the
> VMS  NSF client should handle RMS semantics (forget the actual switches).

Mount command used successfully for several years is:

tcpip mou dnfs1: ALPHA$NFS 
/host=XXX/path="/Volumes/RAID1/WebData"/sys/struc=5/auto/noadf/acp=max_workset=3500 


With the exception of the /acp= added from a recent suggestion
> 
>  My guess is that you have a couple of files on your OS-X file system
> that cause problem when they are fed to the VMS NFS client.

Seems possible - no idea how to track them down as it just started "out 
of the blue" and has survived multiple reboots.

Interestingly, using the /acp command on the mount has stopped the wild 
soft paging, but the ACP still uses 100% CPU.  I have set its priority 
to 0 and it still does it.  To mediate some of the bizarre behavious, I 
now have a small DCL x=x+1 tight loop that I spawn at prioroty 1 - it 
means that other processes get a look in.  But NFS activity still comes 
to a halt until the 100% CPU state is satisfied by (probably) the 
process timing out internally or…..something else in the ACP process.