Subject: | |
From: | |
Reply To: | Computer And Software Enterprises, Inc. |
Date: | Wed, 22 Dec 1999 23:16:24 -0800 |
Content-Type: | TEXT/PLAIN |
Parts/Attachments: |
|
|
In article <[log in to unmask]>,
<[log in to unmask]> wrote:
>First check priority of INETD in the JINETD.net.sys job.
>
>:print jinetd.net.sys
>!job jinetd, manager/?.sys/?
>!run inetd.net.sys;pri=cs
>!eoj
>
>The line #2 should be modified do include ";pri=cs".
Caution! Caution! When you run a process within a job in the CS queue,
and the root CI for the job is in the DS queue (default, and assuming CS
is higher priority than DS and doesn't overlap), if the CS process loops
for any reason, the job will not respond to an ABORTJOB command until
either the root CI is increased to CS priority, or the looping process is
reduced back into the DS queue.
I discovered this wonderful problem when I had a problem with Apache son
processes looping. No new jobs would start. Aborting job did not make
Apache go away. Increased root CI process for Apache job to CS. Apache
job aborted at that time.
The better alternative may be to allow CS job priority. That would allow
the use of PRI=CS on the job statement. However, that would allow any
user to run a job at CS.
This is a problem which needs to be solved before daemons such as inetd
will be reliable on the 3000.
Rick Gilligan
|
|
|