FriarTuck
16th February 2002, 23:02
Does anyone have some insight on the proper way to kill off a user process cleanly.
kill -9 <pid> will kill the process but leave a license unit allocated for the user in licd6.1
kill -TERM <pid> will kill the process and clear the license. Well, that sounds like the answer to my question, but...
If our baan admin comes to me and says there is a user she can't 'kill -TERM' on, will normally kill -9 the oracle Net8 process. I don't think that's such a good idea. So I thought of a different option:
In Oracle, alter system kill session 'sid, serial#'; will work, but is this "safe" under baan? :eek: If it is, then I can use alter system for any time we need to wack a user.
Guntherm
18th February 2002, 10:13
Hi,
You can also use the Baan utility alg6.1 to kill users.
Most of the time we use a 'kill -9 <pid>' to kick out users.
cheers,
Gunther
patvdv
18th February 2002, 15:18
Friar Tuck,
It seems to me that if you do an 'alter system kill session' in Oracle, you will be terminating the bshell's oracle client process rather than the other way around. Based on that I would say that this boils down to a kill -9 for the bshell unless the 'alter system kill session' allows pending transactions to be dealt with correctly.
toolswizard
19th February 2002, 17:27
You should try using bshcmdx.x first. There is a unix script that calls this called alg. I'll attach it for those you don't have it.
yeoea1
21st February 2002, 04:33
The script kill one or all users? The comment in it state tat it is use to kill all user. Pardon me but I do not know how to read the script so only base my question on the comment in the script.
toolswizard
21st February 2002, 04:55
It allows you to select which users or all users. It also allows you to send messages to all logged on users.
NPRao
21st February 2002, 19:51
More info...
bshcmd6.2
Usage: bshcmd6.2 [options] <bshell_pid>
-v : Print version
-p : Show process list
-m : Show memory usage
-d <dbglvl> : Set DEBUG_LEVEL to (octal) <dbglvl>
-k <pid> : Kill bshell process id <pid>
-e : Kill all bshell processes
bshcmd6.2 -k sends the same signal (SIGTERM) as kill -15
FYI on the proper kill sequence:
*be careful how to kill such a Bshell. First try two or three times a
'kill -2', then a 'kill 15' and only if all else fails, try a 'kill -9'.
The first two will trigger the Bshell to execute some emergency code that (amongst other things) de-registers itself at the license server. The 'kill -9' does not leave room for de-registering, resulting in a stranded license.
In the Reger release, the current license mechanism will be
replaced by the BCLM. In this brand-new license manager are many mechanisms to clean-up stranded licenses automatically. So if you are using Reger, you will soon get a version that works with the new license manager that will solve the problems.
norwim
25th February 2002, 15:53
I experienced that the best way to kill a stranded bshell is
bshcmd6.1 -e -u1 -w1 PID (PID = process id of bshell)
The u1 "wakes up" a sleeping bshell
w1 waits for 1 second
You still may meet bshell-processes "immune" to that way, just keep calm and wait. (My theory is that a large sql-statement has been started, only when this is terminated, bshell regains control and can react to the kill-signal)
I don't know about oracle-subprocesses, but it can be fatal to kill a bisam process via kill -9: you for sure terminate the bshell, but with some "luck" tbase stops completely and you have to restart it with a lot of "checking tables"
Of course all the above refers to unix OS, I don't have any idea how NT behaves here.
Norbert
Andy..
28th February 2002, 18:19
depends why you want to kill users, if its to reboot or similar then bschcmd6.x works fine in most cases.
However, if you are looking at saving licences due to idle users you should have a look into CloseIdle.
Reseller for UK and NL is AM Connect www.am-connect.com
mbadillo
11th October 2002, 23:51
If the results of licmon6.2 -w are:
BAAN.PROMESA.COM.EC bw/bi vhaventa.4850 BAAN.PROMESA.COM.EC 11-10-2002 18:35
BAAN.PROMESA.COM.EC bw/bi pgicaja.4860 BAAN.PROMESA.COM.EC 11-10-2002 18:35
BAAN.PROMESA.COM.EC bw/bi aclcredi.5394 BAAN.PROMESA.COM.EC 11-10-2002 19:05
What's the process id (<PID>) wich is referenced in the commadn bshcmd6.2??.
Mario
evertsen
12th October 2002, 00:18
3. Do not cross post. Putting your message/question in one forum should do the trick.
This means you Mario.
Francesco
14th October 2002, 23:55
Originally posted by mbadillo
If the results of licmon6.2 -w are:
BAAN.PROMESA.COM.EC bw/bi vhaventa.4850 BAAN.PROMESA.COM.EC 11-10-2002 18:35
BAAN.PROMESA.COM.EC bw/bi pgicaja.4860 BAAN.PROMESA.COM.EC 11-10-2002 18:35
BAAN.PROMESA.COM.EC bw/bi aclcredi.5394 BAAN.PROMESA.COM.EC 11-10-2002 19:05
What's the process id (<PID>) wich is referenced in the commadn bshcmd6.2??.
Mario
The process ID is the number behind the username. In your example, 4850 would be the PID for vhaventa's bshell.
pokitlok
31st October 2002, 05:58
Dear all,
Should you suggest how to kill the processes of a group of users.
The group may be defined by:
The company they are using
Or, their VRCs / Package Combination
Or, their Package in use
Or, their sessions in use
Thank you for help.
Markus Schmitz
31st October 2002, 11:24
Hi everybody,
If "kill -15" does not work, then bshcmd normally also doesn't work anymore.
Interesting in my experience the following never fails on the client:
In the bw gui open the optione dialog,
activate "Output to console"
a black window will open,
close this window (X)
The whole gui will terminate gracefully ....
This seems to work with Baan IV
We teach all users to use this method, if their GUI hangs. After some time the requests to kill bshells will stop. :-)
Regards
Markus