The Fat Controller v0.0.4 released!
I’m sorry it wasn’t in time for Christmas, I hope everyone managed to have at least a little bit of fun on Christmas day without it – but finally it’s here – version 0.0.4! This version involved the rewriting of the entire logging system for sub-processes, so plenty of testing was needed which wasn’t always compatible with Christmas festivities, (oddly enough).
Download: sourceforge.net/projects/fat-controller/files/
Project web: fat-controller.sourceforge.net
Continuous logging
The main aim for this release was to fully support the “daemonise anything” point of the Fat Controller raison d’etre. Prior to this release, output from sub-processes was collected and then written to the log file only once the process had ended. This was bad for two reasons:
- it only logs output on STDOUT, anything on STDERR is ignored
- no good for long-running daemon processes as nothing is logged until it ends
The Fat Controller now continually monitors STDOUT and STDERR of all sub-processes and immediately writes anything to the log file.
The whole logging system can be re-initialised by sending SIGHUP to The Fat Controller. So if, for example, your log files get deleted due to log rotation, simply send SIGHUP and The Fat Controller will re-open file descriptors to the log files. This can be easily added to your log rotation mechanism.
So, what else is new in v0.0.4? Here’s a brief summary taken from the changelog:
ADDED –run-once
Also to support daemon processes, using the –run-once argument it is possible to tell The Fat Controller to, (cunningly) run a process only once and then end.
ADDED –test-fire
If this argument is specified then The Fat Controller initialises but does not actually run, i.e. it does not daemonise (if it is specified to be in daemonise mode) nor does it run any processes. It is useful combined with the –debug option when testing to check arguments have been correctly read and interpreted.
CHANGED Debug mode
Debug mode is turned on by using the –debug argument when running The Fat Controller. Previously this was added by the init script /etc/init.d/fatcontrollerd if the file fatcontroller.debug was found in the current directory. In addition to turning on debug mode, it also looked for the configuration file in the current directory, and not in /etc/
This has now been substantially simplified. All you need to do to enable debug mode is instead of running the Fat Controller with:
sudo /etc/init.d/fatcontrollerd start
Use:
sudo /etc/init.d/fatcontrollerd debug
My plan is that if no major bugs are found in this release then I will re-release it as v1.0.0 as it will finally be everything that I imagined when I first started this project over a year ago.
I’ve still got plenty more ideas for development and I’m eager to hear any other ideas people may have. Please let me know if you have a great idea or suggestion!
Hello Nick!
I have few comments:
1. information about bad long option
./fatcontroller –cmd
program prints:
Unrecognised option: -
and I don’t know what option
I suggest, change in file dgetopts.c line 264 on
fprintf(stderr, “Unrecognised option: %s\n”, argv[optind-1]);
2. segmentation fault
./fatcontroller -c t1.sh
and I get
segmentation fault
in non daemon mode you don’t check -l option.
I suggest if no -l option, then set log file to /dev/null.
3. information in syslog
./fatcontroller -c t1.sh -l /dev/null
when t1.sh not exists in syslog i found
Dec 30 16:19:59 hp-ux FatController[8067]: Could not replace child process – shutting down child
in my opinion, this information does not say anything
When program run in daemon mode in syslog i found
Dec 30 17:01:53 hp-ux FatController[8406]: Daemon starting up
Dec 30 17:01:53 hp-ux FatController[8406]: Daemon exiting
program runs with pid=8407, and I know nothing about starting daemon with pid=8407
when I kill the proccess 8407 in syslog is correct information
Dec 30 17:06:54 hp-ux FatController[8407]: Daemon exiting
4. -p –path option
In changelog you write that :
-p –path is deprecated
but still using this option in function parseOptions.
As for suggestions on the program, about which I wrote 28.12.2011 ie, reading options from a configuration file, I have it almost complete.
If you are interested to write.
Best regards
Tom
Hi Tom – thanks for this, I’m currently working through it. With regards to point 2, I think the easiest solution is to make -l a required argument. However, if not running in daemon mode it might actually be desirable to tell The Fat Controller not to log anything and instead had STDOUT and STDERR of all sub-processes piped to the STDOUT and STDERR of the Fat Controller process.
Concerning point 4, I left in support for the -p argument so that old configurations will still work. Although, perhaps it can be taken out now so that was a few versions ago.
Great that you’ve been writing something concerning configuration, would be great to see it – I’ll email you so you can send it to me.
Hi Tom. I’ve gone through your list, points 1 and 2 are fixed. As I mentioned previously, specifying a log file is not a required argument. If this is not ideal then it can always be refactored later but simply not starting the sub-process logging service (which would involve either always checking if it is started or for it not to return an error if called when not initialised).
Regarding point 3 – it now gives more information, for example:
“FatController[2025]: Failed to execute command /usr/bin/pytho Error: [2] No such file or directory”
Just committed changes to SVN.