I can not kill the process,nor shutdown for reboot,what caused it?

[frederick van targero_1]

Dear all,

I met embrassed case in one server with oracle installed, I have met for three times, the oracle turn unwork and I cannot shut it down, either I do "killall -9 oracle", nothing affect to the oracle; further more, even if I type "reboot" or "shutdown -r now", the system only echo term signal, and also be there. Until I click the reset button to force it reboot.


ignore the oracle reasons, I doubt what can make a linux can not kill the process which is not in "defunct" yet? and how to reboot the linux forcely even if some process can not be terminated?

thanks in advance

frederick

[Martin P.J. Zinser]

Hello Frederick,

trying to shutdown a database will kill or killall is a <> bad idea. This is a sure ticket to problems with data corruption and data loss. In case of Oracle use shutdown normal or (if required) shutdown immediate from sqlplus with an appropriate account (e.g. system) to shutdown the database. This might also solve your problem with the general shutdown.

Greetings, Martin

[Steven E. Protter]

As the oracle user

sqlplus internal

shutdown immediate

if that does not work

shutdown abort

startup
shutdown immediate

Then look for permission problems on /var fs

SEP

[frederick van targero_1]

thanks martin and steven,

I agree to your suggestion, and Iã did on this way before. However, I can not shut down the oracle once the error occur, for, it will never can not be startup after I type in svrmgrl "shutdown XXX", although it echo shutdown successful, I can also ps its process out as nothing be done. that is why I had to kill oracle process, unfortunately, it also did not worked.

so I call it embarasse, I never thought the linux can not be shutdown before, and, I have to click the button, for even the ms windows is rarely on thus situation.

frederic

[Martin P.J. Zinser]

Hello Frederick,

besides of checking file permissions (as SEP mentioned), you also might want to have a look for open network connections to your database. netstat should report them. I remember having read about hanging shutdown due to problems with network logins to the database.
Which version of Oracle are you running?

Greetings, Martin

[frederick van targero_1]

Hi Martin,

It is linux 7.13, kernel of 2.4.18; and oracle 8.17. for it has been working for a long time, uptime over 250 days. so that I am surprise for this error by myself.

I will check for your recommends. Then, How do you think about the RAM could run out to make similar errors? I have 1000M Ram for the system, and 800M left after the oracle be startup, however, several days later, only tens Megabyte free ram left. does it normal,too?

rdgs,
Frederick



Hello Frederick,

besides of checking file permissions (as SEP mentioned), you also might want to have a look for open network connections to your database. netstat should report them. I remember having read about hanging shutdown due to problems with network logins to the database.
Which version of Oracle are you running?

Greetings, Martin

[dirk dierickx · Accepted solution]

About the only time a process will not react to a kill signal is when it is when it is running in kernel space. that is what could have happened here, something prevented the oracle process to switch back to user space so that it could terminate (probably something io related).

if you wanted to reboot the machine and that would not even work you can always try 'reboot -f'.

[Mark Grant]

Actually, a process that is waiting for some IO to complete will not receive any signal it gets sent and therefore can not be killed.

If you can't re-boot this thing, just turn it off.

[John Garrott]

Frederick,

Have you checked your Oracle alert log? It will usually show the shutdown messages. If you have active network connections to database sessions you will probably see a message about the shutdown process waiting for a session process to complete. Unless the session user logs out, you will need to kill the session (kill -15 will allow Oracle to rollback and minimize the chance of corruption). Once any hanging sessions are killed the database shutdown should resume and complete. If not, use the shutdown abort command in sqlmgrl. Be sure to restart and then shutdown normal or immediate afterwards. If you start the database after a shutdown abort and run without performing a clean shutdown it can result in an unstable database.

You might also check and make sure you are using a terminal release of Oracle and that all recommended Oracle patches are installed. Some earlier releases of 8.1.7 had memory leaks.

John