[jira] Reopened: (DAEMON-65) [daemon] runs as multiple instances, does not use PID file logic

["Dennis Lundberg (JIRA)" <ji...@apache.org>]

     [ https://issues.apache.org/jira/browse/DAEMON-65?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]

Dennis Lundberg reopened DAEMON-65:
-----------------------------------


> [daemon] runs as multiple instances, does not use PID file logic
> ----------------------------------------------------------------
>
>                 Key: DAEMON-65
>                 URL: https://issues.apache.org/jira/browse/DAEMON-65
>             Project: Commons Daemon
>          Issue Type: Bug
>    Affects Versions: 1.0
>         Environment: Operating System: Linux
> Platform: PC
>            Reporter: bernard
>            Priority: Critical
>
> jsvc writes its own pid file but it appears that it does not have a logic that
> secures its own integrity.
> Multiple duplicate processes can be created simply by issuing the same jsvc
> command multiple times. The created processes cannot be killed using the pppid
> file for obvious reasons.
> 1) jsvc should terminate prematurely if it finds its own pid file.
> 2) jsvc should delete its own pid file when killed.
> If 1) and 2) are not acceptable because (hypothetically, because I don't know
> the specifications) the specifications require that the caller incorporates this
> logic, then jsvc should not write a pid file.
> Why do i think so?
> Depending on implementation, the risk of malfunctioning is much higher if the
> pid file is managed across different execution environments.
> One major reason is that these environments are not usually maintained by the
> same person.
> I guess one might try to get a file system lock on the pid file before launching
> the java program.
> Please excuse my ignorance if I am misinterpreting the daemon functionality in
> any way. I have tried to get responses from 3 relevant mailing lists,
> commons-user, commons-dev and tomcat-user, but nobody replied.
> I am not a Linux programmer and I would not be surprised if this kind of
> programming problem (uniqueness of id'd processes on one machine) has a standard
> solution under Linux.
> Because

-- 
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.