Monday, May 21, 2007

Problems changing a BOINC server's name

The following email summarizes a problem that we had when we changed our server's name to make it available outside our firewall.
Solution found:
I wrote:

Okay we changed the name of our server from an internal name to an external one that can be accessed across the Internet. My problem is that bin/start reports:

Entering ENABLED mode
Starting daemons

And nothing happens. "ps aux|grep feeder" only reports the grep instance.

I have changed the server name in the mysql.user table, in /etc/sysconfig/network, in config.xml and I'm sure a few other locations. Restarted the machine, etc. I also verified the permissions on the new log_... and pid_... folders.
Found the problem. The 'stop_daemons' file was not being deleted by bin/start. I did not realize that the config.xml host setting should only be 'www'. Not the FQDN.

The config host has to match the localhost name that is internal to bin/start. This is *not* the same as what is returned by 'hostname'.

This took awhile to find, because no log files were being generated, neither did start report any errors. I only saw it by running feeder under gdb, where the output apparently goes to stdout/stderr instead of (or in addition to) the log file.

--
Ron Parker
UPDATE:
One additional file that needed changed in the SciLINC project directory was html/user/schedulers.txt.

It is also necessary to update the app_version table. The path to the application files on the server is hard coded in the XML data fields of this table. Running bin/update_versions will update it, but it may be necessary to first delete the rows that refer to the old server.