Process Hung

[Frank Dean_1]

Anyone know how to terminate a hung process a hung process without rebooting. OS is OpenVms 7.3 on Vax 6000 hardware. Stop/id does not terminate the process. I believe there is a method using SDA>.

[Volker Halle]

Frank,

terminating a hung process with SDA is not possible. SDA can only read memory locations, but not modify them.

You can use SDA to find out WHY the process is hung. You may then be able to eliminate the reason for the hang. A last resort could be using the DELTA debugger to modify memory locations in system space, but this should only be attempted after thorough analysis, as it can easily lead to system crashes easily.

Let's start with some simple questions:

- what state is the process in ? SHOW SYSTEM
- is is waiting for a LOCK ? SDA> SHOW PROC/LOCK
- does it have busy channels ? SDA> SHOW PROC/CHAN

Volker.

[Jan van den Ende]

Re Volker's remark on DELTA:

In a previous life I have worked on ICL - DME.
They had a program comparable to DELTA, but, just as an extra warning, that program was called
DYNAMITE.
--- you were constantly reminded of the risks involved! I still think it was a much better name than DELTA :-)

fwiw.

Proost.

Have one on me.

jpe

[comarow]

If the porcess is short Bytelimit, and you have Availability Manager on your lan you can save some stuck processes, such as if they are short on bytelimit. It can also fix quorum is the cluster is out of votes. It can fix a lot of things.

Availability manager is FREE, and also warns of many problems. The collector runs on your VMS system, and the controller or display runs either on a VMS system or a PC (on the same lan).
However, it is better to run it on a PC as it uses JAVA and eats a ton of memory on a VMS system.

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

[Peter Quodling]

Define "Hung" in this context? IS the process in a resource wait state (i.e. RW???) . If so, SDA may help in finding what it is being caused by, which is a starting point on fixing it.

In SDA, try a show proc/index of the rogue process. Look for 0/nnn fields, indicating something has run out of a resource.

then try a show proc/channel and look to see what channels are busy for that process.

After that, a lot depends on what you get, so report back...

q

[Frank Dean_1]

Thanks to the responses; the troubleshooting tips using SDA mentioned in your responses were used and point to the application. The programmers are looking at the code.

[Frank Dean_1]

Thanks to the responses; the troubleshooting tips using SDA mentioned in your responses were used and point to the application. The programmers are looking at the code.