On Nov 23, 2015, at 4:36 PM, Oscar Jofre <oscar@jofre.com> wrote:
I've been setting up mariadb database and radius: radiusd.conf max_requests = 250000 max_servers = 256
That is way too high, unless you're running 10M users.
database: max_connextions = 500 (because before 141 reach tha max value)
Fix your database so it isn't slow.
I keep testing 12.000 PPoE connection with 5 minutes timeout and 1 minute interim.
That's ridiculous. 1 minute interim updates are *way* too short. Set it for 5-10 minutes at least. And what "timeout" did you set? Session-Timeout? If that's at 5 minutes, it's also ridiculous. Set it to at least an hour.
To reach 144.000 new rows per hour.
Which is ridiculous. No wonder your database falls over. You have configured your server so that it will break. Don't do that. Use *reasonable* defaults.
Now on radius log I got:
1) Every minute I got
Mon Nov 23 22:07:01 2015 : ERROR: (3057208) detail: ERROR: Couldn't open file /var/log/radius/radacct/88.17.150.241/detail-20151123: Failed to open file /var/log/radius/radacct/88.17.150.241/detail-20151123: No such file or directory Mon Nov 23 22:07:01 2015 : ERROR: (3057209) detail: ERROR: Couldn't open file /var/log/radius/radacct/88.13.76.4/detail-20151123: Failed to open file /var/log/radius/radacct/88.13.76.4/detail-20151123: No such file or directory
You misconfigured something locally. That message doesn't appear in the default configuration.
I've changed privileges
find /var/log/radius -type d -exec chmod 755 {} \; find /var/log/radius -type f -exec chmod 644 {} \;
But still the same. The funny thing is that the detail file is growing, I mena there is updating info every second.
The default configuration writes accounting packets to the detail file.
2)
I don’t understand how I have to handle this spares Info .. I've changed min_spare_servers = 30 and max_spare_servers = 400 but always this info on log:
Mon Nov 23 22:23:02 2015 : Info: rlm_sql (sql): Need 4 more connections to reach 226 spares
Fix your configuration so it's not using ridiculously short timeouts.
During this test the radacct whas performing between 1.900 and 2.100 inserts per second
Because you set ridiculously short timeouts. Dont' do that.
Thanks for all the help I really appreciate it !
You've been told repeatedly to not set such short timeouts. You're still doing it. You've already been told *why* it's a problem. Instead of fixing the timeouts, you keep using the same short timeouts. And then ask (again) why there's a problem. We've told you why it's a problem. Stop asking questions, and follow instructions. Alan DeKok.