Urgent help, radius is crashing due to Child is hung for request

[bbnlrad at bbnldevices.in]

Hi,
I am new to radius, hence please forgive me if i am asking very basic
stupid questions.

We have a production box having radius 2.2. it is talking to a mysql
server 5.1.
The broadband clients are coming in through mikrotik router 6x

>From past a month a month we are seeing that the radius is getting into
issues intermittently and does not respond. we keep seeing the following
errors during that time.
ue Nov  3 22:25:59 2015 : Info: WARNING: Child is hung for request 12221
in component <core> module <queue>.
Tue Nov  3 22:25:59 2015 : Info: WARNING: Child is hung for request 11643
in component <core> module <queue>.
Tue Nov  3 22:25:59 2015 : Info: WARNING: Child is hung for request 12795
in component <core> module <queue>.
The connections on the db side is around

In the radacct table we see around 8 to 9k users online.
the mysql database is running on a dual core with around 1% loadaverage
and the user cpu is just getting consume around 12%. Based on this i do
presume the db is running fine.

A small obseravtion from my team is that if the users crsossed around 10k
the the issues starts.

Can somebody please do help me out as to what configuration needs to be
done in the radius server for a large volume we have around 60k customers,
what internal tweaking needs to be done on the radius server side.
Also please help me out as to what recommendation needs to set on the db
side too.

one the radiusd.conf file down below is what i see which is un commented
max_request_time = 30
cleanup_delay = 5
max_requests = 2048
thread pool {
	start_servers = 5
	max_servers = 32
	min_spare_servers = 3
	max_spare_servers = 10
	max_requests_per_server = 0
}


sql.conf has
	num_sql_socks = 200
	connect_failure_retry_delay = 60
	lifetime = 0
	max_queries = 0
	nas_table = "nas"

the db has
symbolic-links=0
innodb_stats_on_metadata=0

key_buffer=128M
myisam_sort_buffer_size=128M
sort_buffer_size=32M

innodb_buffer_pool_size=5120M
innodb_log_file_size=256M
innodb_file_per_table
innodb_flush_log_at_trx_commit=2
innodb_flush_method=O_DIRECT


when we start seeing the errors in the adius logs, i have tried to login
to the mysql from the radius server, it takes time to login.

The machine is a quad core with 8 gb ram

Your help is greatly appreciated.


Thanks
Arun