| T.R | Title | User | Personal
Name | Date | Lines |
|---|
| 1935.1 | Also AUDIT doesn't help | MLNCSC::BORGINI | | Wed Mar 19 1997 04:17 | 13 |
| Hi,
any news about this problem?
The customer has already activated the AUDIT, collecting alarms
like breakin, logfailure and file_access, but when the MAIL error
has occurred there weren't useful informations in the display
of the command SHOW AUDIT.
This problem is very random, last occurence was last sunday
in two different moments.
Any suggestions?
Regards,
Roberto.
|
| 1935.2 | | EPS::VANDENHEUVEL | Hein | Sun Mar 23 1997 01:25 | 58 |
|
> %MAIL-E-NOMSG, message number 007e828a
> and it is necessary to reboot to delete this failure.
As you read in topic 1329, this is simply a file locked error.
Sure, reboot will clear that, but it is a mighty big hammer.
Your customer deserves better support that he is currently getting.
You just have to go chase the owner of the file lock.
$SHOW DEV/FILE is possibly the easiest way to find out which
process has the file locked. ANAL/SYS... SHOW LOCK... is
an alternative. Also check out a bunch of (old) tools to list
who is holding a file lock.
A normal reason for a mail file lock is auto-compress after an auto-purge.
Just switch of auto-purge to get control back.
>There aren't info useful in ERROR.LOG and OPERATOR.LOG; they have tried
>ANA/DISK/REP of the system disk, without help.
Ofcourse there is nothing useful there. There is no serious error,
there is no failure, there is no corruption. This is likely to be
standard operating procedures.
> They have a defragmentation product named DISKSKIPPER running, actually
Now _that_ conceivably could lock a file while defragging.
Switch it off, or run during controlled hours to learn more.
> I've checked the fragmentation index of INDEXF.SYS with DUMP/HEADER
> command and it is good (only 5 counts in the Map area).
Totaly, utterly, unrelated. Really, you are wasting the customers
time and resources. Please escalate locally with someone that
know someting more about vms.
>I've suggested them to insert in the startup of each system the logic
>MAIL$SYSTEM_FLAGS = 7 (before it wasn't defined), but the error has returned
>this sunday.
See above. Please call for help to avoid Digital from looking too silly.
> Now I think, like last chance, to restore an old image backup
> before first appearance of the MAIL error, but the customer has serious
> problems to plan a stop of his cluster, and would other possible tests.
See above. Dont waste the customer's time by random activities.
The file is locked. Just go try and find out by who.
It could be a simple compress. it could be a poorly implemented
defrag product. In the latter case you may be able to help your
customer best by teaching them how to avoid mail-file fragmentation
through $CONVERT/FAST/NOSORT/FDL=xxx where xxx is an optimized
FDL with reasonable allocation/extent.
|
| 1935.3 | See DECamds lock contention display... | XDELTA::HOFFMAN | Steve, OpenVMS Engineering | Mon Mar 24 1997 14:14 | 14 |
| : Your customer deserves better support that he is currently getting.
eg: if this is a serious customer question, log an IPMT.
: You just have to go chase the owner of the file lock.
: $SHOW DEV/FILE is possibly the easiest way to find out which
: process has the file locked. ANAL/SYS... SHOW LOCK... is
: an alternative. Also check out a bunch of (old) tools to list
: who is holding a file lock.
DECamds is a good tool for chasing around lockholders and lock
contention products... (DECamds is part of OpenVMS, though it
is seperately installed.)
|
| 1935.4 | We will try to disable DISKSKIPPER | MLNCSC::BORGINI | | Tue Mar 25 1997 10:57 | 19 |
| Hi Hein,
thanks for your suggestions. Thanks also to Steve, I will consider
the possible use of DECamds for this case.
My customer will try to disable DISKSKIPPER, to check what happen without
this tool running. In effect, this defragmenter software runs at precise
hours on all ten disks, locking in that moment the files.
When the error appeared, SHOW DEV/FILE displaied 000000.
However, reading the note 1329.1, I had not clear that the error was
surely %RMS-E-FLK, if you substitute the facility code 015 at this
message you will obtain %LIB-E-INVTYPE, invalid LIB$TPARSE state table entry.
With regard to our service, the customer doesn't complain about it;
he is instead satisfied about our help, also thanks the use of this tool.
Moreover I've not facilities to find other VMS experts here, many of them
don't work any more in DEC and others are difficult to find, so I use
this tool to have many precious suggestions.
Regards,
Roberto.
|