radutmp is not big, it has only 700kB However radwtmp has almost 700MB pet

------------ Původní zpráva ------------ Od: Marinko Tarlać <mangia81@gmail.com> Předmět: Re: Freeradius stopped to work Datum: 20.2.2012 13:37:16 ---------------------------------------- radutmp file became to large...

But to be sure, check you radius in debug mode...

/usr/local/var/log/radius/radutmp

On 20.2.2012 13:25, dorje2007@seznam.cz wrote:

...

Hi

after more then year my freeradius 2.1.9 stopped to work

in log I have a lot of the following info:

....Feb 20 13:11:55 radius radiusd[12006]: WARNING: Child is hung for request 35 in component accounting module radutmp. Feb 20 13:11:56 radius radiusd[12006]: WARNING: Child is hung for request 34 in component accounting module radutmp. Feb 20 13:11:56 radius radiusd[12006]: WARNING: Child is hung for request 51 in component accounting module radutmp. .... Feb 20 13:10:45 radius radiusd[12006]: WARNING: Unresponsive child for request 134, in module radutmp component accounting Feb 20 13:10:47 radius radiusd[12006]: WARNING: Unresponsive child for request 132, in module radutmp component accounting Feb 20 13:10:48 radius radiusd[12006]: WARNING: Unresponsive child for request 136, in module radutmp component accounting

and so on

When i restart the radius, it start to work, then after a few minutes goes down

Any idieas ?

thanks

pet - List info/subscribe/unsubscribe? See http://www.freeradius.org/list/users.html

hi yes, i;m using accounting, but i don't know for what is the radwtmp file responsible thanks

------------ Původní zpráva ------------ Od: Alan Buxey <A.L.M.Buxey@lboro.ac.uk> Předmět: Re: Freeradius stopped to work Datum: 20.2.2012 13:46:43 ---------------------------------------- Hi,

...

However radwtmp has almost 700MB

are you using it - ie any of the features that require it? If not, then turn off the calls to it in accounting etc -

alan - List info/subscribe/unsubscribe? See http://www.freeradius.org/list/users.html

------------ Původní zpráva ------------ Od: Fajar A. Nugraha <list@fajar.net> Předmět: Re: Freeradius stopped to work Datum: 20.2.2012 14:05:19 ---------------------------------------- 2012/2/20 <dorje2007@seznam.cz>:

...

hi yes, i;m using accounting, but i don't know for what is the radwtmp file responsible thanks

Are you using radwtmp for accounting?

If you store accounting data in sql or detail file, or don't know what radwtmp is, chances are you don't need radwtmp.

As Alan suggested, comment-out radutmp in your sites-available/default (or whatever other virtual server you might use)

-- Fajar

Thank you I've commented-out the line "unix" in accounting section a removed the radwtmp file Will see what will happen : ) thanks pet

Hi,

unfortunately , radius wend down again. The log is not very precise:

Feb 20 14:22:44 radius radiusd[12700]: WARNING: Child is hung for request 988 in component module . Feb 20 14:22:44 radius radiusd[12700]: WARNING: Child is hung for request 990 in component module . Feb 20 14:22:44 radius radiusd[12700]: WARNING: Child is hung for request 992 in component module . Feb 20 14:22:44 radius radiusd[12700]: WARNING: Child is hung for request 994 in component module . Feb 20 14:22:45 radius radiusd[12700]: WARNING: Child is hung for request 1025 in component module .

lots of records = live accounting has reached a block

i didn't comment out the "radutmp" line, because i need to have a n accounting

yes - but HOW do you DO accounting? you say you need to have radutmp - but how do you use that file - are you sure you use it and not eg SQL accounting? if you have got to a stage where the accounting requirements cannot be met in real time, then you need to look at migrating to using eg detail file 'just out of live time' accounting - or moving the accounting off the live auth server and using eg decoupled accounting, or proxy to home server buffered accounting.. or even move to a more efficient method such as buffered-sql . however, the issue may also be caused by some other local issue - such as network contention (if you proxy requests), resource starvation (host is doing more than just the radiusd and doesnt have enough resource - be that CPU or disk access - to do its job), or, if virtualised, the host may not have all it needs because some other virtual host is stealing the cycles now ..or you've finally encountered a bug. does this happen with latest release (2.1.12)? alan

...

Feb 20 14:22:44 radius radiusd[12700]: WARNING: Child is hung for request 988

in component module .

That means it's getting hung somewhere in the server core. That's usually because the system is receiving more requests than it can handle.

You need to fix it so that it can handle high volumes of traffic. Upgrade the machine, add DB indexes, do less processing work per packet.

Hi, how i can fix it ? The server is virtual , 2GB RAM, 1x 3.33 GHZ CPU The server is doing almost nothing, load averaga less then one: Here is the "top" command: top - 17:49:36 up 25 days, 6:24, 1 user, load average: 0.20, 0.15, 0.10 Tasks: 79 total, 2 running, 77 sleeping, 0 stopped, 0 zombie Cpu(s): 0.0%us, 0.0%sy, 0.0%ni, 88.0%id, 12.0%wa, 0.0%hi, 0.0%si, 0.0%st Mem: 2075492k total, 1261764k used, 813728k free, 153996k buffers Swap: 2096472k total, 60k used, 2096412k free, 996640k cached Also as i said, nothing happend more then one year, and suddenly today it has started to crash Thanks