Apple Enterprise - NeXTanswers Support Archive
Enterprise
[Index]
[Download]


Search NeXTanswers for:

Q: Help! My machine has slowed to molasses. Even executing ls in a terminal takes forever. A ps shows the process numbers in the PID table are in the thousands--even soon after reboot. There are 3 defunct processes owned by root with priority of -1 which can't be killed. When trying to launch memory hungry applications on my low memory (RAM memory) machine I get ``insufficient memory'' errors. What can I do?

A: The key symptom of this problem is the defunct or ``zombie'' processes shown in the PID table. A zombie process occurs when a child process dies after its parent has died, and no wait() has been done for the child. Once the child process dies, it is inherited by init and owned by root. These processes are not consuming any system resources, except for one process table slot. They are not using memory or CPU time because they are done executing and therefore cannot be killed. When a UNIX system exhibits this type of problem (spawning lots of zombies) then some system process (often a getty) is trying to start, is failing, and is re-trying endlessly.

Good places to look for causes of such problems are the following system files (all in /etc):

rc
rc.boot
rc.local
rc.swap
crontab
ttys
gettytab

Perform a checksum (using sum) on the suspect file on your system and compare to the checksum of the same file on your Software Release master. A difference in the two numbers does not necessarily indicate an error--you might have intentionally changed something--but it does mean the file should be more closely scrutinized, especially if you don't think you've changed the file.

For example, the first several lines of the /etc/ttys file on a stock, unmodified Release 1 through Release 3 disk looks like this:

#
# name getty type status comments
#
# If you do not want to start the window server by default, you can
# uncomment the first entry and comment out the second.
#
# console "/usr/etc/getty std.9600" NeXT on secure
console /usr/lib/NextStep/loginwindow NeXT on secure window=/usr/lib/NextStep/WindowServer onoption="/usr/etc/getty std.9600"
ttya "/usr/etc/getty std.9600" unknown off secure
ttyb "/usr/etc/getty std.9600" unknown off secure
ttyda "/usr/etc/getty D9600" unknown off
ttydb "/usr/etc/getty D9600" unknown off
ttyp0 none network
ttyp1 none network

The man page for ttys explains that the flag on and off in the fourth column specify whether init should execute the command given in the second field. It's an error to specify on if the second field contains none. The secure further qualifies an on to allow root to login to this line. (See more about the ``secure'' flag below.)

Specifying on so that someone may log in over the network tty line causes the performance degradation described above. The network entries should never specify on. In other words, an /etc/ttys file which contains the following entries is an error:

ttyp0 none network on secure
ttyp1 none network on secure
ttyp2 none network on secure

If this is the case, either change the on entry to off and reboot, or restore the file from a Release disk and reboot. As an alternative to rebooting, after modifying the ttys file, you can type in a Terminal as root:

rhino-22# kill -HUP 1

This tells the init process to reread the ttys file.

Warning: specifying off and secure on ttyp0 (for example) is permissible but not advisable. This would allow users to rlogin or telnet as root to the machine on pseudo port 0 and can present a security risk. See Chapter 16, ``Security,'' in the Release 2 NeXT Network and System Administration or Chapter 14, ``Security,'' in the Release 3 NeXTSTEP Network and System Administration manual for more information on security in general.

QA592

Valid for 1.0, 2.0, 3.0, 3.1



OpenStep | Alliances | Training | Tech Support | Where to Buy