[...]only checks whether the process is present (by pid).
A little hint:
If using chrooted like virtualization technologies it might be counterproductive to rely on process names or /proc/<PID>/exe, ...
ZBXNEXT-2152 is somewhat related
Fixed in development branch svn://svn.zabbix.com/branches/dev/ZBX-5558 .
The fix includes the following:
- startup error messages are no longer hidden (as per this issue);
- a stale PID file that points to a non-existing process is now removed (as per
ZBX-5379); - the script was adapted to work under /bin/sh (as per
ZBXNEXT-1889).
Successfully tested
(1) regarding shutdown.sh - we just send kill signal but do no verify/wait for it to shut down. something like shutdown.sh && startup.sh is almost guaranteed to fail.
we probably will not want to touch these scripts for a while, so it might be a good idea to deal with this now.
asaveljevs RESOLVED in r43186. The shutdown script now waits for the gateway to stop.
wiper CLOSED
(2) another suggestion - we might want to include a commented out line that would enable jmx (w/o ssl, simplest possible) for startup.sh.
asaveljevs Added an example to startup.sh in r43189. RESOLVED.
wiper CLOSED
(3) Not sure whether this is an issue, but in case we have no permissions to create a PID file, the gateway starts normally with the following output from startup.sh:
$ ./startup.sh ./startup.sh: 1: ./startup.sh: cannot create /zabbix_java.pid: Permission denied cat: /zabbix_java.pid: No such file or directory Zabbix Java Gateway did not start
Do you think this is a problem?
wiper I think starting and telling it did not is wrong. Also in this case shutdown script will give false information about gateway not running.
asaveljevs Added your suggestion of touching the PID file before starting the gateway. RESOLVED in r43197.
wiper CLOSED
Successfully tested
Fixed in pre-2.0.12 r43228, pre-2.2.3 r43229, and pre-2.3.0 (trunk) r43230.
i got really curious why this works as intended :
STDOUT=`$COMMAND_LINE & echo $! > "$PID_FILE"`
that is, why subshell, initiated by ampersand, is actually waited upon before backtick subshell exits.
for example, bash manpage says :
If a command is terminated by the control operator &, the shell executes the command in the background in a subshell. The shell does not wait for the command to finish, and the return status is 0."
thus one might expect the command to be executed immediately and then it's output spilling out later, potentially. or, to have a simpler example :
a=`sleep 3 & echo 1`
here those 3 seconds are waited upon before backtick shell returns.
the reason is stdout not being disconnected in command substitution (both for java process and sleep 
) - if stdout would be closed, backtick subshell would terminate.
listed here : http://stackoverflow.com/questions/16874043/bash-command-substitution-forcing-process-to-foreground and also thanks to kurahaupo from #bash
(4) Added "Startup and shutdown scripts for Java gateway no longer hide error messages on startup, detect stale PID files, and should now work in /bin/sh." to:
- https://www.zabbix.com/documentation/2.0/manual/introduction/whatsnew2012?&#daemon_improvements
- https://www.zabbix.com/documentation/2.2/manual/introduction/whatsnew223?&#daemon_improvements
asaveljevs Please take a look. RESOLVED.
martins-v A slight problem here that "no longer" could be understood as "no longer detect", too. That's not intended I believe, or is it?
asaveljevs You are correct, Mārtiņš. That was not intended.
martins-v I split the sentence into two, to avoid ambiguity. If that's alright, this could be closed.
asaveljevs Thank you! CLOSED.