77nafets

[Log in to get rid of this advertisement]

Hi there!

I have the problem that I want to identify a process reliably after some time (i.e., after weeks or even months) to check whether it is still running or not.

This however seems rather difficult: I cannot check the PID, since PIDs are only unique during the lifetime of the process.

The maximum PID value is 32767, so when more processes are launched during uptime, the PID counter wraps around, starting from the beginning again. You can easily verify this with a shell script:

#!/bin/bash

if test -r "$0"; then
echo "pid=$$";
"$0"&
fi;

Name this 'wrap'. When launched, it will run until you remove/rename its file. While running, it repeatedly restarts itself, occupying a new PID. The output looks something like

> ./wrap
[...]
pid=32765
pid=32766
pid=32767
pid=300
pid=301
pid=302
[...]

I also cannot use a combination of PID and command, since the same program may have been relaunched, accidentally occupying the same PID. This is not very likely, but not impossible. Just see the 'wrap' script above.

Especially since the process I want to watch is to be relaunched repeatedly, it is very likely that it will have a PID assigned that it used before.

Is there something like a "totally unique" PID that's never reused during uptime? I guess a 64-bit number should be enough.

Any suggestins?

Thanks!

tronayne

You might look at PID, PPID, STIME and TIME in combination; have a look at what ps -ef shows you or ps -elf.

openSauce

I don't understand what it is you want. If "the process" is relaunched repeatedly, each relaunch will be a separate process. Assuming the first one has come to an end, you already know the answer to your question without checking anything - no, the process is not still running.

77nafets

Yes, cool idea. I might use a combination of PID and starttime (e.g. see proc(5)). Thank you.

---

No.

I start a process 'foo' in the background (using bash syntax).

> foo&
[1] 4321

It's easy to test whether a process with such a PID exists:

> kill -0 4321 && echo alive
alive

> kill -0 9999 && echo alive
bash: kill: 9999 - No such process

Now assume 'foo' terminates.
The PID counter wraps around.
Another 'foo' process is started, accidentally having the PID 4321 assigned.

> kill -0 4321 && echo alive
alive

Even looking in the process table will tell me that it's 'foo' running with PID 4321. But The original process died.

---

Any other suggestions?

Thanks,
Stefan

openSauce

Ah I see. Sorry, misunderstood you. Don't really know a good way to do this, but a quick-and-dirty way might be something like:

The parentheses start a new shell in which to execute the commands, which (I'm hoping) would protect the $starttime variable from being unset in the intervening months. tronayne's way is probably better.