I found the update that fails the launch, but I don't understand it. I'm not friendly to Python so I'll detail step by step what I did. The revison that doesn't work for me is r3474. To test it I cleaned a folder, installed the 3473 revison by following the Roy Hooper's guide (http://blog.royhooper.ca/2007/07/07/installing-the-darwin-calendar-server-on...). It works for my current revison (3443). Revision 3473 works. I cleaned the directory, (rm -r /usr/local/test_calendar/*) and installed r3474 with the same procedure. Launching fails. Launch fails at the same step I reported in my first mail. So I admitted the r3474 is the cause of the problem, or reveals a problem. I watched the changes to r3474 : http://trac.calendarserver.org/changeset/3474/ This is a change to the number of processes. I'm not an expert in Python and programing generally speaking. I know the GetNCPU function doesn't work on FreeBSD 7.0, as reported in the logs. So I tried to bypass the change by setting 'ProcessCount == 1' and 'MinProcessCount == 1' in config.py to match the current number of processes in my version. But it doesn't work. I tried with 'ProcessCount == 1' and 'MinProcessCount == 1', same result. I'm now blocked since I don't know what to test. I'll be glad if someone can help. Thank you, Bastien Semene Bastien Semene a écrit :

I solved the problem by using r3443. I'll try to find what commit is exactly failing for me and the reason behind this.

Bastien Semene a écrit :

...

Hi everyone,

I'm using CalendarServer since some months on FreeBSD 7.0 and it worked fine. I didn't updated it since a while and it seems the r3599 (or older) fails to launch but I have no log about it.

I made a new install to test the last revision. Creating a new directory and fetching the last revision through SVN. I did the changes to make it work on FreeBSD : - modifying bash's bin path - autorizing root to launch memcached (this was in cluster.py in my previous install, so according to Trac it is a really old revision I worked on).

This the launch log :

calendar# ./run Using python as Python

exec python /usr/local/calendar/Twisted/bin/twistd -n caldav -f /usr/local/calendar/CalendarServer/conf/caldavd-dev.plist -o ProcessType=Combined 2009-01-20 11:05:56+0100 [-] Log opened. 2009-01-20 11:05:56+0100 [-] twistd 2.5.0+rUnknown (/usr/local/bin/python 2.5.2) starting up 2009-01-20 11:05:56+0100 [-] reactor class: <class 'twisted.internet.selectreactor.SelectReactor'> 2009-01-20 11:05:56+0100 [-] [calendarserver.tap.caldav.CalDAVServiceMaker#error] Unable to detect number of CPUs: getNCPU not supported on freebsd7 2009-01-20 11:05:56+0100 [-] [calendarserver.tap.caldav.CalDAVServiceMaker#info] 0 processors found. Configuring 4 processes. 2009-01-20 11:05:56+0100 [-] [calendarserver.tap.caldav.CalDAVServiceMaker#info] Adding memcached service 2009-01-20 11:05:56+0100 [-] twistedcaldav.accesslog.AMPLoggingFactory starting on "'logs/caldavd.sock'" 2009-01-20 11:08:19+0100 [-] Received SIGINT, shutting down. 2009-01-20 11:08:19+0100 [-] (Port None Closed) 2009-01-20 11:08:19+0100 [-] Main loop terminated. 2009-01-20 11:08:19+0100 [-] Server Shut Down.

I stopped the process at 11:08:19. I put the log level in caldavd-dev.plist at "debug" level, but there's nothing in the error log.

I tried to find informations on google without results. I'm regularly reading the threads here but find nothing related to this problem.

At this point I don't know where to find more information about this problem, so any idea or redirection is welcome. Thank you,

Bastien Semene _______________________________________________ calendarserver-users mailing list calendarserver-users@lists.macosforge.org http://lists.macosforge.org/mailman/listinfo.cgi/calendarserver-users

_______________________________________________ calendarserver-users mailing list calendarserver-users@lists.macosforge.org http://lists.macosforge.org/mailman/listinfo.cgi/calendarserver-users

Ok it works. I made a new caldav working directory at r3474 and applied your fix by hand, the server worked succesfully. I updated it to HEAD version and it worked succesfully also. Now I'll consider patching the deployed server to the HEAD revision. I'll find a way to patch (locally) the server to autodetect correctly the number of processors on FreeBSD as I'm loosing performances I think. I'll say when I found the answer to this problem. Cyrus Daboo a e'crit :

Hi Bastien,

--On January 22, 2009 11:36:29 AM +0100 Bastien Semene <bsemene@cyanide-studio.com> wrote:

...

This is a change to the number of processes. I'm not an expert in Python and programing generally speaking. I know the GetNCPU function doesn't work on FreeBSD 7.0, as reported in the logs. So I tried to bypass the change by setting 'ProcessCount == 1' and 'MinProcessCount == 1' in config.py to match the current number of processes in my version. But it doesn't work. I tried with 'ProcessCount == 1' and 'MinProcessCount == 1', same result.

I'm now blocked since I don't know what to test. I'll be glad if someone can help.

I just committed a fix in r3607. Please give that a try.

NB To get it working without that fix you need to edit your caldavd.plist file and set the "ProcessCount" key value to 1 (or more) there. That value will override what is in config.py as a default.