Unable to stop process

[Sue van Wyk]

Hi,
On rx7620, OpenVMS 8.2-1 a telnet session connection is in a disconnected state.

24F3746B TNA2821: (disconnected)

This process however is using 100% CPU time.

0 25 50 75 100 + - - - - + - - - - + - - - - + - - - - + 24F3746B IIS1035207 100 ▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒▒

I am unable to stop the process. I have tried "Stop/ID" "Stop/proc" "Suspend process" nothing stops it.

"SHOW SYS" looks like this:
24F3746B IIS1035207 CUR 3 1 272191 1 21:57:48.37 10997 703

Please, is there anyone who experianced this before, and maybe found a way of stopping the process without rebooting the server.

Thanks

[Ian Miller.]

If you have CMKRNL priv then can you
$ ANAL/SYSTEM
then at the SDA> prompt do
SHOW PROC/ID=24F3746B

and post the result here.

[Sue van Wyk]

See attachment for ana/sys output of process.

[Ian Miller.]

To see where the process is
In SDA do
READ/EXEC
SET PROC/ID=24F3746B
SHOW PROC/CHAN ! to see what the 1 open file is
SHOW STACK ! repeat a couple of times to see if it varies.

[Sue van Wyk]

I did a show proc/channels and the output is attached.
The show stack does change every time I show it. Don't know if that helps.

The executable shown in the attached output from show channels is the program the user was busy with when the session disconnected.
This executable is share by many users at this time.

[Hein van den Heuvel]

Almost 2 days of CPU already huh? Boy those Itaniums are fast :-)

Did you have a peek with $SHOW PROC/CONT ?

Anyway, It smells like a bug. This being an Itanium I would not waste any time here. Go straight to HP support, make it an urgent call.. just in case it is a bug they may want to fix it before 8.3 ships or to learn that perhpas it has been fixed.

[ Hopefully you are able/willing to take a crash with dump if that data is needed. ]


fwiw,
Hein.

[Sue van Wyk]

You said it Hein.....

When I do: sh proc/cont/id=24F3746B
I get: F-SUSPENDED, process is suspended

Thanks
I have logged it with HP.
Just thought someone came across a similar problem.

[Ian Miller.]

"show proc/cont/id=24F3746B
I get: F-SUSPENDED, process is suspended"

That's because the process is marked for delete.

[Peter Barkas]

If the process is hung on a BG device you can disconnect the device using system command tcpip disconnect device bgxxx. See also tcpip sho dev bgxxxx/fu.

[Volker Halle]

Sue,

please consider to issue a couple of

SDA> SHOW CALL/SUMM

commands against the looping process. This should show the call stack in a abbreviated format (better to read than a SHOW STACK).

I'm assuming the process seems to loop in some rundown code.

You could try a SET PROC/PRIO=0 to reduce the impact of the looping process to the rest of the system.

Volker.

[Jan van den Ende]

Sue,

it may be a small consolation, but we also know the phenomenon.
Alpha, VMS 7.3-2.

$ SET PROC/ID=.../PRIO=0 converts it into an alternate NULL proces until convenient to reboot (provided it does not hold any other resources needed by others).

Proost.

Have one on me.

jpe

[Volker Halle]

Sue,

another good tool to collect information about the looping process is PCS$SDA (the SDA PC Sampling extension):

$ ANA/SYS
SDA> PCS ! to get some help information
SDA> PCS LOAD
SDA> PCS START TRACE/PID=
SDA> PCS STOP TRACE
SDA> PCS SHOW TRACE/STATISTICS
SDA> PCS UNLOAD

This will list the PC values sampled from the looping process sorted by descending no. of samples for individual PC values.

This will tell you, in which routines the process is looping.

Volker.

[Sue van Wyk]

Thanks for the advise guys, but it seems like this is a bug and there is no way to stop this process other than to reboot the system. I rebooted....
This thread can be closed.