1000's of defunct processss

[Grayh]

I have observer that these defunct processes are increasing by time... due to which the users are getting disconnected...

.
.
.
oracle 8803 22583 2 06:35:11 ? 0:00
oracle 8787 22583 1 06:35:00 ? 0:00
oracle 8863 22583 1 06:35:52 ? 0:00
oracle 8862 22583 2 06:35:51 ? 0:00
oracle 8799 22583 1 06:35:01 ? 0:00
oracle 8870 22583 2 06:36:02 ? 0:00
oracle 8808 22583 1 06:35:23 ? 0:00
oracle 8926 22583 2 06:38:15 ? 0:00
oracle 8871 22583 1 06:36:13 ? 0:00
oracle 8872 22583 1 06:36:14 ? 0:00
oracle 8928 22583 1 06:38:32 ? 0:00
oracle 8881 22583 2 06:36:24 ? 0:00
oracle 8907 22583 2 06:38:04 ? 0:00
oracle 8900 22583 2 06:37:42 ? 0:00
oracle 8921 22583 1 06:38:14 ? 0:00
oracle 8929 22583 1 06:38:33 ? 0:00
oracle 8927 22583 1 06:38:25 ? 0:00
oracle 8930 22583 1 06:38:43 ? 0:00
oracle 9009 22583 1 06:40:11 ? 0:00

.
.
.
# ps -ef | grep 8930
oracle 8930 22583 1 06:38:43 ? 0:00
root 19967 19732 0 10:53:40 pts/0 0:00 grep 8930
# ps -ef|grep 9009
oracle 9009 22583 1 06:40:11 ? 0:00
root 19978 19732 1 10:54:00 pts/0 0:00 grep 9009

Here the parent process is root and I cannot Kill it...

Is there something that is generating these zombie processes...

[Ganesan R · Accepted solution]

Hi Grayh,

It looks like all defunc processes are oracle related.
In your example root is not the parent process for zombies. PPID is the parent process id.

Use #ps -fp to know which process is responsible for the zombie processes.

You cannot kill zombies. Reboot is the only way to get rid out of this.

[Srimalik]

the parent for every process is a process with PID 22583
please see which process have PID 22583.

It seems that it is not calling one of the wait family calles.

[Steven E. Protter]

Shalom,

Only a reboot will kill the zombies.

The oracle application or the cause of the disconnect needs to be dealt with before the system grinds to a halt.

SEP

[Grayh]

ps -fp 22583
UID PID PPID C STIME TTY TIME COMMAND
oracle 22583 1 0 15:18:33 ? 35:21 ora_j000_HP01


And for the users they get the below error when trying to connect to Oracle DB

" ora-12514 tns listener could not resolve service_name given in connect descriptor "

[Srimalik]

> Only a reboot will kill the zombies.

Bouncing the database will also cleanup the zombies.


>The oracle application or the cause of the disconnect needs to be dealt with before the system grinds to a halt.

Yes, before reboot you need to make sure that these zombies are the actual problem which are causing connection failures.

>" ora-12514 tns listener could not resolve service_name given in connect descriptor "

This type of error comes when oracle sid is not exported properly at the client side or the oracle listner configuration is faulty. But, As I have not used oracle for many days I am not sure about it.

[OldSchool]

"I have observer that these defunct processes are increasing by time... due to which the users are getting disconnected..."

That is exactly backwards. The defuncts are, more than likely, the result of the disconnect, not the other way around.


"# ps -ef | grep 8930
oracle 8930 22583 1 06:38:43 ? 0:00
root 19967 19732 0 10:53:40 pts/0 0:00 grep 8930
# ps -ef|grep 9009
oracle 9009 22583 1 06:40:11 ? 0:00
root 19978 19732 1 10:54:00 pts/0 0:00 grep 9009

Here the parent process is root and I cannot Kill it... "

Nothing in the above would indicate that the "parent process" is root. you might consider running:

UNIX95= ps -fHp 22583

to see who the parent is.

fix the connect/disconnect issue and the defuncts should be resolved.

[Dennis Handly]

>Is there something that is generating these zombie processes.

Yes, the zombie master's (ora_j000_HP01) poor coding. It seems it doesn't do error handling properly for that connection issue.

>Ganesan: You cannot kill zombies. Reboot is the only way to get rid out of this.
>SEP: Only a reboot will kill the zombies.

This is incorrect. Killing the zombie master will kill the zombies.

[Grayh]

Thanks everyone