[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

ps fax shows 100's of <defunct> processes in Sarge



I just did an apt-get upgrade from woody to sarge and it glitched.

There were two errors:  a bug in the nano install script and a bug in the tkx8.3 install.  These cause the apt-get upgrade to abort.

I was able to fix the nano problem quite simply and I have reported this.

For the tkx8.3 problem I did apt-get install tkx8.3  This worked.  I suspect at this point apt erased the information it needed for the upgrade.

After apt installed all the dependancies for tkx8.3 it declared the upgrade was complete.  In fact - it had not installed kde for instance and there may be other problems left over from its abortion.


However the system is working - almost.  

when I do ps fax I see 100's of <defunct> processes listed from deamons and apps like: Privoxy, xmms, eog, artsd, firefox, kpdf.  I assume these processes are defunct because the parent has not wait()ed its child and hense the child becomes a zoombie.  However this did not occur in woody and it is a consequence of the upgrade.

Thus, I am asking for assistance on how to categorise this bug.  It is serious.



The information I have found on defunct processes states that the zoombie is simply kept in the process table until something wait()s it.  This documentation says this is harmless.  This documentation is incorrect.  This is not harmless.  Eventually the application in question will clog up and not be able to fork() and what it does at that point depends on the error handling in the application.  Privoxy for instance burps with an error message sent back to the browser.  I do not yet know what the other apps do.

When I killed privoxy then of course all the zoomies under that parent pid were destroyed because privoxy's parent knew what to do with them and wait()ed them to death.

kill of course does not work because these zoombies are already dead.


I am running on the old woody kernel:   2.4.18-bf2

I will upgrade this too - soon - but I have noted that if this is a problem then the upgrade scripts do not identify this and they did not upgrade the kernel.  In this case the upgrade scripts are broken in yet another way.

That would IMHO be yet another bug.

It appears the woody to sarge upgrade process is not debugged and I have to admit surprise because there is no indication in the debian website that this is not a smooth process.

Terrell Larson



Reply to: