There are two fundamental problems with the status command.

1) It reports a tomcat instance as running when the instance isn't
running and reports pids which do not belong to the instance.

2) It does not respect the CATALINA_PID file variable which is set in
the per instance config file and instead hardcodes the pid file to
/var/run/${NAME}.pid

Problem 1 occurs when there are multiple tomcat instances on the
system. The status command checks for the presence of
/var/run/${NAME}.pid, if the pid file is not present it is supposed to
indicate the instance is stopped. But instead if /var/run/${NAME}.pid
is not present it uses pgrep to search the process table for any
java process running with the TOMCAT_USER uid and gid. When there are
multiple instances pgrep will return a list of pids for each
instance. However, none of those instances are the instance whose
status we're checking. Thus it reports the instance is running, but
that is not correct, the instance is not running but other instances
are and it's those instances whose pid's are reported.

Problem 2 occurs whenever the instance config does not use the same
pid file default encoded in the initscript. For example we group our
pids in a subdirectory of /var/run specific to the application.

Ideally we would like to use the lsb status command and not implement
the status command in our initscript. However the lsb status command
takes as one of it's arguments the process name. However we don't know
the process name because it's set in the wrapper via set_javacmd and
that process name is not known in the initscript. Attached is a patch
which reimplements the status command following the LSB guidelines,
and respects the CATALINA_PID variable.