topic Re: Process crashes while backup is activ in Operating System - OpenVMS

Process crashes while backup is activ

Dario Karlen — Wed, 12 Oct 2005 01:11:21 GMT

we have a process who's getting information from other nodes all the time and writes the data in a .DAT-file. Of course we have to do a backup of this dat-file. but when backup is activ it locks the file and the process crashes, although the parameter /ignore=interlock is set. even if i exclude this dat-file und try to backup the rest of the directory the process crashes sometimes, is it possible that the backup locks the whole directory??? what can i try next?

Re: Process crashes while backup is activ

labadie_1 — Wed, 12 Oct 2005 01:42:44 GMT

Could you add a few details, architecture (Vax, Alpha, Itanium), Vms version (6.2, 7.3-2...) the quotas of the username running the process (mc authorize sh xxx), post the command procedure, to get an idea of what it does...

When the process crashes, you get some more information, could you post it ?

Re: Process crashes while backup is activ

Karl Rohwedder — Wed, 12 Oct 2005 03:01:14 GMT

Dario,

you did not mention if your program is high level or just DCL, but in either ways the OPEN and WRITE statements normally provide back status information, e.g. in DCL WRITE/ERROR=...
and you should be able to get some usefull errormessages.
If you .DAT files stays opened by your program all the time, the BACKUP may produce an error about an open file, but your program should not crash.

regards Kalle

Re: Process crashes while backup is activ

Dario Karlen — Wed, 12 Oct 2005 04:04:45 GMT

this is the errormessage:
%FOR-F-OPEFAI, open failure unit 114 file DSA2:[WAL.DATA]SHIFT_STAT.DAT;
-RMS-E-FLK, file currently locked by another user
%TRACE-F-TRACEBACK, symbolic stack dump follows
image module routine line rel PC abs PC
0 FFFFFFFF8045F5DC FFFFFFFF8045F5DC
DEC$FORRTL 0 0000000000029B30 000000007C60DB30
DEC$FORRTL 0 00000000000371E4 000000007C61B1E4
PL PL PL 7342 000000000000EF08 000000000003EF08
0 FFFFFFFF87C5F3F4 FFFFFFFF87C5F3F4
WAL job terminated at 12-OCT-2005 07:26:07.44
Accounting information:
Buffered I/O count: 40784124 Peak working set size: 8720
Direct I/O count: 1376433 Peak virtual size: 170880
Page faults: 822 Mounted volumes: 0
Charged CPU time: 0 00:23:13.19 Elapsed time: 1 21:05:22.33
this is the backup command:
$ Backup/rewind/journal=list_tag.bjl/record/noassist/ignore=inter -
dsa2:[wal.data]*.*;*-
/exclude=(devlog.dat,ui.dat,pkt*.*,quen*.*,shift_stat.dat) srveb1$mkd500:data_tag.bck/save_set/rewind/ignore=label

Re: Process crashes while backup is activ

Dario Karlen — Wed, 12 Oct 2005 04:11:00 GMT

OpenVMS V7.2-1
Alpha DS10

it crashes on this statement:
7342 OPEN (STAT_LUN,FILE='WAL_PCU_SHIFT_STATS',
7343 2 ACCESS='DIRECT',
7344 2 ORGANIZATION='RELATIVE',
7345 2 RECL=PRO_DATA_REC_LENGTH/4,
7346 2 STATUS='UNKNOWN',
7347 2 SHARED)

Re: Process crashes while backup is activ

Karl Rohwedder — Wed, 12 Oct 2005 04:20:47 GMT

"-RMS-E-FLK, file currently locked by another user"

says it all, another process has the file open.

Since you excluded the file from the backup command, are you sure, that it is the BACKUP process, which has the file opened?

A SHOW DEVICE /FILES/NOSYSTEM on the disk may be of help.

Sidenote: you use the /IGNORE qualidier twice on your backup command, I think the 2nd takes precedence and the /IGNORE=INTER is ignored, try /IGNORE=(INTERLOC,LABEL).

regards Kalle

Re: Process crashes while backup is activ

Dario Karlen — Wed, 12 Oct 2005 04:43:03 GMT

the process is running 24 hours a day and when it gets a message from another host it tries to open that file. when backup is activ and a message is comimg from a host the process crashes.
this happens ONLY when backup is activ. i had a look with SHOW DEV /FILES/NOSYSTEM but no other process uses this file.

i will try with the /IGNORE(INTERLOCK,LABEL)

Re: Process crashes while backup is activ

Karl Rohwedder — Wed, 12 Oct 2005 04:47:00 GMT

The /IGNORE=INTER will not help here, because it tells BACKUP to backup files which are locked by OTHER users, but does not mean to allow other user to open files which are opened by backup.

Btw. habe you checked that the EXCLUDE works, i.e. is the file really missing from the saveset?

regards Kalle

Re: Process crashes while backup is activ

Dario Karlen — Wed, 12 Oct 2005 06:07:58 GMT

that's clear with the /ignore=inter parameter.

but! right, the file IS in the save set! i don't know why, but i can see the file in the list_tag.bjl. something is wrong with the parameter dsa2:[wal.data]*.*;* /exclude=(shift_stat.dat)
any idea?

Re: Process crashes while backup is activ

Karl Rohwedder — Wed, 12 Oct 2005 06:36:01 GMT

ODS-5 volume and case sensitivity enabled?

regards Kalle

Re: Process crashes while backup is activ

Hein van den Heuvel — Wed, 12 Oct 2005 06:49:06 GMT

Several thoughts:

1) Just teach the program to deal with the error. On open failure, wait X seconds and try again. Abort only after Y retries

2) If the program is running 24/24 then just have it keep the file open (shared)?

3) If the program can not keep it open, perhaps because it is restarted from a dcl process loop, then have the process keep it open with DCL: $OPEN/READ/SHARE=WRITE x x.dat

4) If the process can not keep it open, then have an other process keep it open. This will prevent Backup from obtained exclusive access and will allow the program to re-open it in compatible (shared) mode.

hth,
Hein.

Re: Process crashes while backup is activ

faris_3 — Wed, 12 Oct 2005 06:59:57 GMT

Strange.

On OpenVMS 8.2,

With a backup active (with /ignore=interlock) i am able to open the file from a fortran with the same options than you all the time.

If i remove the /ignore=interlock, i get
the same error.

Re: Process crashes while backup is activ

Doug Phillips — Wed, 12 Oct 2005 10:24:41 GMT

Kalle explained it. /ignore is a command qualifier (not an input or output qualifier) and the second usage caused the first to be ignored.

Re: Process crashes while backup is activ

faris_3 — Wed, 12 Oct 2005 11:12:04 GMT

Thanks. I had not red it carefully enough.

And without the IGNORE=INTERLOCK, Backup
locks the file.

Re: Process crashes while backup is activ

Doug Phillips — Wed, 12 Oct 2005 11:21:35 GMT

The surprise is that most of us didn't think /ignore=interlock would have this effect!

Re: Process crashes while backup is activ

Dario Karlen — Thu, 13 Oct 2005 01:58:07 GMT

it works when the process has keep the file open.

but!
i don't want to do a backup of these files and exclude them. but when i have a look at the backup journal file, the file is listed. what's wrong with my backup command?

$ Backup/rewind/journal=list_tag.bjl/record/noassist/ignore=(inter,label) -
dsa2:[wal.data]*.*;*-
/exclude=(devlog.dat,ui.dat,pkt*.*,quen*.*,SHIFT_STAT.DAT,arb_plan.*,-
tap.dat),-
dsa2:[wal.s7data]*.*;*/exclude=(statlog.dat,binstat.dat),-
dsa2:[wal.progs]*.*;*,-
dsa2:[wal.dario]*.for;*,-
dsa2:[wal.leo]*.for;* -
srveb1$mkd500:data_tag.bck/save_set/rewind

Re: Process crashes while backup is activ

Karl Rohwedder — Thu, 13 Oct 2005 02:11:53 GMT

Dario,

it is the same problem as with the /IGNORE quailifier, you use the /EXCLUDE quailifer twice in your command, I fear only the last one is used. Try to put all files to exclude in one /EXCLUDE statement.

regards Kalle

Re: Process crashes while backup is activ

Dario Karlen — Thu, 13 Oct 2005 03:52:02 GMT

damn, such a weak programming :)
sorry, but i'm a newbie and very very happy about the help here.
that was it, i can't use the exclude qualifier twice.

last question: i give the qualifier /record. so i want that in the next backup only chaged files will be backed up. but every time i start backup EVERY file is backed up, why? shouldn't it back up only changed files?
thanks
dario

Re: Process crashes while backup is activ

Karl Rohwedder — Thu, 13 Oct 2005 04:05:55 GMT

You have to add the qualifier /SINCE=BACKUP.

Please read the help on this HELP BACKUP /SINCE.

regards Kalle

Re: Process crashes while backup is activ

Ian Miller. — Thu, 13 Oct 2005 04:06:42 GMT

to pick up files changed since last backup then add /SINCE=BACKUP to your command.