Issues with Freeradius crashing after a sighup

[Alex Sharaz]

Hi all,

I've inherited a pair of Freeradius servers running Vsn 2.10 and have build a new server around the 2.2 source code. All of these servers exhibit the same problem in that after a SIGHUP to reload their configuration files the sometimes crash.

Firstly the 2.1 servers

We have 2 of them configured to support our wired and wireless auth user base. Each server has a primary auth function ( wired or wireless) and acts as a backup for the other server)

These are running on an old Debian OS and make use of the Freeradius versions available through the apt-get package manager. Configuration wise  everything uses password files and all logs are written to a local hard disk.  We don't use SQL or AD or any other systems in the authentication or accounting process.

password files are updated every 15 mins and are followed by a "service freeradius reload" command to bring them on line. 

At least once a day the freeradius daemon will crash just after the "reload" command. The normal logfiles  (see below) just show the following with no indication of why the process crashed. Cfashes happen randomly on both servers, although the server handling the wireless network crashes more frequently than the one handling the wired network.

Fri Feb  8 00:05:03 2013 : Info: HUP - loading modules
Fri Feb  8 00:05:03 2013 : Info:  Module: Reloaded module "attr_filter.post-proxy"
Fri Feb  8 00:05:03 2013 : Info:  Module: Reloaded module "attr_filter.pre-proxy"
Fri Feb  8 00:05:03 2013 : Info:  Module: Reloaded module "attr_filter.access_reject"
Fri Feb  8 00:05:03 2013 : Info:  Module: Reloaded module "attr_filter.accounting_response"
Fri Feb  8 00:05:03 2013 : Info:  Module: Reloaded module "pap"
Fri Feb  8 00:05:03 2013 : Info:  Module: Reloaded module "files"
Fri Feb  8 00:05:03 2013 : Info:  Module: Reloaded module "accounting_log"
Fri Feb  8 00:05:03 2013 : Info:  Module: Reloaded module "auth_log"
Fri Feb  8 00:05:03 2013 : Info:  Module: Reloaded module "reply_log"
Fri Feb  8 00:05:03 2013 : Info:  Module: Reloaded module "pre_proxy_log"
Fri Feb  8 00:05:03 2013 : Info:  Module: Reloaded module "post_proxy_log"
Fri Feb  8 00:05:03 2013 : Info:  Module: Reloaded module "york_passwd"
Fri Feb  8 00:05:03 2013 : Info:  Module: Reloaded module "landb_device_info"
Fri Feb  8 00:05:03 2013 : Info:  Module: Reloaded module "switch_vlan_info"
Fri Feb  8 00:05:03 2013 : Info:  Module: Reloaded module "sql_log"
Fri Feb  8 00:05:03 2013 : Info:  Module: Reloaded module "suffix"
Fri Feb  8 00:05:03 2013 : Info:  Module: Reloaded module "mschap"
Fri Feb  8 00:05:03 2013 : Info:  Module: Reloaded module "mschap_default"
Fri Feb  8 00:05:03 2013 : Info:  Module: Reloaded module "detail"
Fri Feb  8 00:05:03 2013 : Info: Loaded virtual server <default>
Fri Feb  8 00:05:03 2013 : Info: Loaded virtual server inner-tunnel
Fri Feb  8 00:05:03 2013 : Info: Loaded virtual server eduroam

Freeradius version 2.2 - wireless server

The 2.2 server was compiled from source on an Ubuntu 12.04 LTS VmWare server and has a slightly different configuration. Configuration files are used for MAC based authentication and for some standard users such as the university of york eduroam health check test account. For 802.1x authentication I use a back end AD system and authenticate all our "real users" against AD. Configuration files for MAC based authentication RADIUS clients and test users are generated once a day and the system is reloaded at midnight every day.  

The configuration used on this server is based upon the "template" one provided by UKERNA  for their UK eduroam user base.

This server can run for a couple of weeks before it crashes.

I know I should run the daemon with the -X option and dump the output to a file, but given the random nature of these crashes, I'm not sure I'll have enough disk space to just run in debug mode and  collect all the logs.

Anyone else seen serve crashes on a reload?

Rgds
Alex