[Info-vax] Nonexisting "Owner process" of a device.

[Jan-Erik Soderholm]

Den 2017-07-07 kl. 09:17, skrev Simon Clubley:
> On 2017-07-06, Jan-Erik Soderholm <jan-erik.soderholm at telia.com> wrote:
>> Den 2017-07-06 kl. 19:08, skrev Simon Clubley:
>>>
>>> Have you checked the status of the underlying network connection ?
>>>
>>> Has the underlying network connection completed the TCP timeout
>>> sequence or is it still transitioning through the various TCP
>>> timeout states ?
>>>
>>
>> If I knew where to look I might answer that one...
>>
> 
> $ tcpip netstat -na
> 
>> We simply create TNA devices pointing at some IP/port
>> and then the applications use them as terminal devices.
> 
> The question here is if the underlying TCP connection is staying
> around for some reason.
> 
> Simon.
> 

I might add also that, before doing the "tcpip /delete" and "tcpip /create"
to "refresh" the TNA device, the scripts does a STOP on the application
that uses the port (probably the process ID still left as owner).

In another thread it was said that STOP doesn't run the exit handlers.

Right, I found traces from that process. It was the process (a interactive
VT-application) that uses that TNA port to write to a label printer).
See parts from the Rdb monitor log below.

  6-JUL-2017 16:11:56.89 - Received user attach request from 0007547C:1
   - process name xxxxxxxx, user xxxx
   - image name "xxxxxxxxxx"
   - database name "xxxxxxxx"
   - 20 global buffers allocated; 15160 free out of 20000
   - sending normal user attach reply to 0007547C:1
...
...

  6-JUL-2017 16:13:37.60 - Received user image termination from 0007547C:1
   - database name "xxxxxxx"
   - abnormal user termination detected
   - %SYSTEM-S-NORMAL, normal successful completion
   - database monitor created Database Recovery process RDM_RB72_09F82
   - user termination suspended until recovery ready
$

The process was probably STOP'ed by the batch job while it still had
waiting I/O, probably due to that we had lost contact with the printer.

Wasn't there some "hack" to clear the "owner" field of a device?

I think we should look for some other way to "stop" a process from
our restart scripts then to simply do a STOP on it. The problem is
that, if the TNA port has stuck, the process will also be stuck
in a QIOW...

Jan-Erik.