STOP/ID for process in LEF

[Ruslan R. Laishev]

[quote]
Next time you get into this state, run the attached program, and paste the output here..
[/quote]
Thanks, it's not a virus ? :-)

[Stanley F Quayle]

> Have you process priority greater than zero and less than 16?

Priority zero is valid. In VMS before V5.0 there was a "NULL" process. The "idle" state is now lower than priority 0. There is no need to avoid priority 0. (Even VMS has its urban legends...)

> Thanks, it's not a virus ? :-)

I suppose you could trust me. Alternatively, run it from an account that has no privileges but access to the process status (same UIC or privilege WORLD).

[Ruslan R. Laishev]

[quote]
Priority zero is valid. In VMS before V5.0 there was a "NULL" process. The "idle" state is now lower than priority 0. There is no need to avoid priority 0. (Even VMS has its urban legends...)
[quote]

SYSGEN's PIXSCAN also here...



[quote]
I suppose you could trust me. Alternatively, run it from an account that has no privileges but access to the process status (same UIC or privilege WORLD).
[/quote]
Ok. I'll.
Is there something what I can extract from SYSDUMP.DMP with SDA ?

[Stanley F Quayle]

> Is there something what I can extract from SYSDUMP.DMP with SDA ?

Probably. But I wouldn't know how -- I'm no SDA expert.

You've already identified that the process is in the RMS layer, so SDA probably won't give you much information.

I'd start with your program's design, and check for everywhere that does I/O (I'm assuming that I/O is your problem).

You might also do a "SHOW DEVICE/FILE D", which will show you all the open files on each disk. Filter for the process ID of the process in question. Might be interesting/useful.

[Willem Grooters]

Second thought: could be a mailbox as well. Does your program read a mailbox, if writer doens't exist (or doesn't write) the programn will wait in LEF.

On SDA:
1. get all privileges (you'll need WORLD anyway, SYSPRV and CMKRNL are likely)
2. run ANALYZE/SYSTEM
3 SHOW SUMMARY will give you a list of current programs. Locate the one that you want to examine, take the value in the second column and issue command:
4. SET PROCESS/INDEX=
5. SHOW PROCESS and look at Direct IO, Buffered IO and Buffered IO Bytecount.
6. SHOW PROCESS/LOCK to see if any lock on a device exists. If not, issue:
7. SHOW PROCESS/CHANNEL, and look for channels "Busy" (If a lock exists on a channel, this may hang your SDA session...)

That's the one it's waiting for (most likely). Give us the details.

Willem

[Ruslan R. Laishev]

Hi !
Some more info is in attachment lef.log.

Thanks.

[Ruslan R. Laishev]

Stanley,

$ zz :==$DISK$SYSMAN:[LAISHEV]109428.EXE
$ shs http

Alpha/VMS V7.3 on node DTV1 3-JUN-2004 15:19:44.75 Uptime 7 04:09:25
Pid Process Name State Pri I/O CPU Page flts Ph.Mem
2E800254 HTTPd:80 LEF 9 15935789 0 01:49:19.31 26182 50784 O
$ zz
PROCMON Process Monitor Utility V2.1
_Process name or /ID=pid: 2E800254
%SYSTEM-W-NONEXPR, nonexistent process
$

[Ruslan R. Laishev]

Hi All!
Now I have a process wich is "Hung in LEF", DEC Amds shows " blocking resource APPENDER"...
What is a resources named "APPENDER" ?

[Ruslan R. Laishev]

So, who interesting by possible problem in VMS/Alpha 7.3-2 can look into the two attached log files.

[Stanley F Quayle]

> PROCMON Process Monitor Utility V2.1
> _Process name or /ID=pid: 2E800254
> %SYSTEM-W-NONEXPR, nonexistent process

This looks like a process ID. You have to say:

$ zz /id=2E800254

since a process could have a name of "2E800254", there's no way for the program to choose without the /ID= switch.

[Ruslan R. Laishev]

3-JUN-2004 16:20:29.60 P R O C M O N V2.1 Node: DTV1
Process Name: HTTPd:80 PID: 2E800254 Index: 412
Username: INET_HTTP Owner: UIC: [375,20]
Mode: Other JobType: Detached Created: 27-MAY-2004 11:13:30.88
$1$DUA126:[HT_ROOT.AXP.][000000]HTTPD_SSL.EXE;137
+------------------------------------------------------------------------------+
| CPU time: 0 01:49:19.40 State: LEF Status: 00040411 |
| |
| Quota Remain/Limit WS Default: 5920 AST's enabled: K SU |
| AST 495/512 WS Quota: 11840 AST's active: U |
| BIO 505/512 WS Extent: 770048 EF cluster 0: C0000001 |
| BYT 422592/422592 EF cluster 1: 80000000 |
| DIO 510/512 WS Max Siz: 64016 EF Wait Mask: F7FFFFFF |
| ENQ 1827/2000 WS Lst Siz: 68320 Cur/Limit SubProc: 2/64 |
| FIL 464/512 WS Gbl Pgs: 15888 Cur/Base Priority: 9/4 |
| PGFL 443568/512000 WS Prc Pgs: 34928 Peak Virtual Size: 269760 |
| TQ 126/128 WS Cur Siz: 50816 Cur. Virtual Size: 259216 |
| |
| Buffrd I/O: 12393499 Authorized privileges: 00000000 5310B825 |
| Direct I/O: 3542290 Default privileges: 00000000 5310B825 |
| Page Faults: 26184 Current privileges: 00000000 10108000 |
+------------------------------------------------------------------------------+

[Ruslan R. Laishev]

3-JUN-2004 16:22:32.60 P R O C M O N V2.1 Node: DTV1
Process Name: HTTPd:80 PID: 2E800254 Index: 412
Username: INET_HTTP Owner: UIC: [375,20]
Mode: Other JobType: Detached Created: 27-MAY-2004 11:13:30.88
$1$DUA126:[HT_ROOT.AXP.][000000]HTTPD_SSL.EXE;137
+------------------------------------------------------------------------------+
| CPU time: 0 01:49:19.40 State: LEF Status: 00040411 |
| |
| Quota Remain/Limit WS Default: 5920 AST's enabled: K SU |
| AST 495/512 WS Quota: 11840 AST's active: U |
| BIO 505/512 WS Extent: 770048 EF cluster 0: C0000001 |
| BYT 422592/422592 EF cluster 1: 80000000 |
| DIO 510/512 WS Max Siz: 64016 EF Wait Mask: F7FFFFFF |
| ENQ 1827/2000 WS Lst Siz: 68320 Cur/Limit SubProc: 2/64 |
| FIL 464/512 WS Gbl Pgs: 15888 Cur/Base Priority: 9/4 |
| PGFL 443568/512000 WS Prc Pgs: 34928 Peak Virtual Size: 269760 |
| TQ 126/128 WS Cur Siz: 50816 Cur. Virtual Size: 259216 |
| |
| Buffrd I/O: 12393499 Authorized privileges: 00000000 5310B825 |
| Direct I/O: 3542290 Default privileges: 00000000 5310B825 |
| Page Faults: 26184 Current privileges: 00000000 10108000 |
+------------------------------------------------------------------------------+

[Willem Grooters]

Quite some locks on resource "APPENDER".
Each of them ia s child-lock - parent is by RMS, as far as I could see on HTTPDLOG:

What I encountered - and may give a hint - is that ALL the locks on HTTPDLOG have a blockingAST (to be executed when it blocks another, higher state lockrequest) and none of the childlocks on APPENDER have - except for a few.

Take a look to the following devices, what is their counterpart:

0210 7FF40400 00000000 Busy BG27663:
0230 7FF40440 00000000 Busy BG27672:
0250 7FF40480 00000000 Busy BG27674:
0210 7FF40400 00000000 Busy BG27663:
02A0 7FF40520 00000000 Busy

(use TCPIP SHO DEV/FULL, output to some file for easier examination)

MBA5735:
02E0 7FF405A0 00000000 Busy MBA5739:
0320 7FF40620 00000000 Busy MBA5743:
0390 7FF40700 00000000 Busy MBA5747:

(SHO DEV /FULL MB, output to some file for easier examination)

You'll find the mailbox names, then you may get some idea who should read, or write

Willem

[Stanley F Quayle]

> AST's active: U

It's "stuck" in a user-mode AST. You might want to check your AST routines...

[Willem Grooters]

Just another thing: the process seems to be waiting for EFN 27 (if I counted well).

You may need the sources to find out where this flag is used....

(Anyone knowing the source could help as well. Am I right this is the WASD or OSU webserver?)

Willem

[Ruslan R. Laishev]

efn = 27 is reserved by RMS. :-(

-----------------------------------------
2.7 Reserved Event Flags

RMS uses system-reserved event flags to synchronize its internal operations. RMS reserves event flags 27, 28, 29, and 30 for possible use; in addition, event flag 31 is used to specify a "do not care" event flag for asynchronous processing.
-----------------------------------------

[Ruslan R. Laishev]

Hi Williem!

Yes, in this case the process is WASD HTTP Server by Mark G. Daniel.

[Willem Grooters]

Well Ruslan,

Given this info, I have a feeling that just one IO on HTTPDLOG is waiting for completion, and JUST ONE lock on resource APPENDER seems to block this IO.
Given the name, it _may_ be the resource for extending this file. Having read comments on WASD, it wouldn't surprise me if IO is done using the bare principle RMS IO - so the program will have to take care of locking itself probably.
You may want to contact Mark Daniel for more information and a REAL solution, it this problem persists.

[Ruslan R. Laishev]

Williem,
it's not an user-application specific problem. In other case do I'm need to think thah WASD can DoS-ing the VMS by simple RMS I/O operations ?!

[Ruslan R. Laishev]

Moreover, the resource named APPENDER is used by RMS itself:

$ search [.rms...]*.* appender

******************************
$2$DUA100:[000000.V73.RMS.LIS]RM0SHARE.LIS;1

1 463 ! The append lock may be acquired from another appender.
3 524 Q_RESNAM [0] = %CHARCOUNT ('APPENDER');
3 525 Q_RESNAM [1] = UPLIT ('APPENDER');
45505041 0000 .ASCII \APPENDER\
$


If you have the VMS source kit you can take a look at:

2 512 ! Have we ever taken out an append lock on this file?
2 513 !
2 514 IF .SFSB [SFSB$L_EOF_LOCK_ID] EQL 0 THEN
2 515 BEGIN
3 516
3 517 ! No, so we need to set up the resoure descriptor and
3 518 ! acquire the lock for the first time. First get a Null mode lock
3 519 ! using LCK$M_EXPEDITE to avoid lock manager wait queue lockout.
3 520 ! This lock should be granted synchronously, so no completion or
3 521 ! blocking AST is specified and the EFN used is IMP$C_ASYQIOEFN.
3 522 ! If not granted synchronously then bugcheck.
3 523 !
3 524 Q_RESNAM [0] = %CHARCOUNT ('APPENDER');
3 525 Q_RESNAM [1] = UPLIT ('APPENDER');

And so on...

[labadie_1]

Ruslan

You should install Amds or Availability Manager, which will give you good in formation about this problem

check

http://h71000.www7.hp.com/openvms/products/availman/

regards

Gerard

[Ruslan R. Laishev]

labadie, I run DECAmds for the past several years. This facility is not supposed to "solving end-user problem with hung in LEF".

[labadie_1]

Ruslan

sure, but it can say
users Dupond holds the APPENDER resource that the following 28 users are waiting for

or a similar message.

regards

Gerard

[Ruslan R. Laishev]

labadie,
sure, you can say that :-)

APPENDER resource - it's a resource used by RMS internaly and maintained by RMS itself, why do you think that it need to search a user "Dupond" who block this resource?
DEC Amds displayed message like "HTTPD:80 blocking resource APPENDER" and what?

[Martin P.J. Zinser]

Hello Ruslan,

is it always WASd having this problem? If yes, what version of WASd are you running?

Greetings, Martin