Help with new Integrity

[Aaron Lewis_1]

I have 2 standalone rx6600's with VMS8.31h1
Both should be identical. One one system I am getting this error:

%DCL-E-OPENIN, error opening SYS$SYSDEVICE:[MYDIRECTORY]QUELIST.DAT; as input
-RMS-E-ACC, ACP file access failed
-SYSTEM-F-EXBYTLM, exceeded byte count quota
*** Exiting due to error encountered

FM21$ dir/date/size quelist.dat

QUELIST.DAT;3 exceeded byte count quota
QUELIST.DAT;2 exceeded byte count quota
QUELIST.DAT;1 exceeded byte count quota

Total of 3 files, 0 blocks.

My account has both: UIC 1,4 & bypass, I only get this on one system, the other one is fine.
And all setting for my account on both systems are identical

If I log out and back in, it’s works fine. It is reproducable.

[labadie_1]

Hello

The help/message says

help/message EXBYTLM



Facility: SYSTEM, System Services

Explanation: The requested operation failed because the byte count quota of
the process is not large enough. This can occur if excessive
concurrent buffered I/O is outstanding, if a large number of
mailboxes is created, or if a large number of windows needs to
be created to completely map a file.

Failure on window creation can occur on a $CRMPSC, $CREATE_
GFILE, $CRMPSC_FILE_64, or $CRMPSC_GFILE_64 system service
call, the DCL command RUN, or an Access or Create of a file.
In this case, this message indicates that the specified file
is very fragmented or the byte limit quota of the process
should be increased.


What is the bytlm of the 2 processes ?

Check the bytlm and bytcnt (in fact JIB$L_OR
G_BYTLM) using the procedure I posted at
http://forums12.itrc.hp.com/service/forums/questionanswer.do?admit=109447627+1212579054833+28353475&threadId=712900

What is the value of the system parameter MAXBUF on both systems ?

[Aaron Lewis_1]

The bytlim from the UAF is 900000. The com file size only 14 blocks, and quelist.dat is only 25, so I don't think fragmentation would be the issue.

[labadie_1]

What is the value of MAXBUF on both systems ?

[Aaron Lewis_1]

sorry about that, it is 8192 on both systems

[labadie_1]

and how much bytlm do you still have ( use the procedure of my first reply) ?

May be the login or sylogin does different things on the 2 nodes, and you have just a little bytlm lef t on a node ?

[Hein van den Heuvel]

As Labadie request, at least a $SHOW PROC/QUOTA, and better still some getjpi bytlm/bytcnt data in the failing situation and working situation would be helpful.

As you have a recent OpenVMS version (8.3) you can also look into the process from a different window using : $SHOW PROC/CONT/ID
Now hit "q" for the dynamic quoate display.
See how it changes through the life of the process.

It must be building up open files, outstanding IO and so on.

Is the problem specific to one (set of) file(s) as suggested or does the DIR/DATE fail on all files once it fails?

If it is one (set) of files, then try other things: Can you do DIR/FILE. Probably yes, as that does not open the file. Can you do DUMP/HEAD/BLO=COUNT=0?...
And uh.. what does DIR/FULL or DUMP/HEAD/BLOCK=COUNT=0 look like when it succeeds?

Good luck!
Hein.

[Hoff]

Was this SYSUAF and related copied from elsewhere?

Have you verified this under a correctly-provisioned SYSTEM username?

Here are a few quota-related discussions:

http://64.223.189.234/node/141
http://64.223.189.234/node/226
http://64.223.189.234/node/49

Various memory-related process quotas can potentially need to be doubled or better from (good) Alpha values, and even the older OpenVMS Alpha quotas have seen increases into the V8 range. And existing quota values can quite easily be somewhere between marginal and entirely insufficient. And there are specific recommendations around increasing quotas on existing SYSTEM users even on OpenVMS Alpha releases.

OpenVMS I64 physical memory tends to best be double or triple that of a reasonable and equivalent OpenVMS Alpha memory configuration, too, at least in terms of executable code size. (And being faster and usually more capable than most any Alpha systems, more is better.)

The "so I don't think fragmentation would be the issue" is an odd approach toward debugging. Years spent in debug has taught me that certain sorts of assumptions can lead me badly astray. Here, even with these files, I'd still check disk fragmentation. Sure, it might be a waste of time, but -- if the fragmentation had been checked -- it's quite and easy and a minimal waste of time, and it's also work that allows fragmentation to be ruled out as a potential trigger.

And FWIW, your account should not be [1,4], unless you are SYSTEM. That's bad practice from the perspective of accountability; it's the sort of thing that (good) auditors will catch, too.

In general, do load current ECOs for OpenVMS I64 V8.3-1H1, too. DIRECTORY blowing up is either a system-level error, or the process quotas are way out of whack.

The BYPASS privilege is only centrally relevant to overriding object access and object protections, and not to quotas.

Stephen Hoffman
HoffmanLabs LLC