Server crashes

[Amr el-Saeed]

after the server finishes  starting the mysql connections, it prints 
that error   Error: FATAL: Thread create failed: Cannot allocate memory 
, and starts to connect to mysql again and the error again and so on

Amr el-Saeed wrote:
> i'm running Linux version 2.4.21-51.EL 
> (brewbuilder at ls20-bc2-14.build.redhat.com) (gcc version 3.2.3 20030502 
> (Red Hat Linux 3.2.3-58))
>
>
>
> Amr el-Saeed wrote:
>> Dear Alan,
>>
>> Thanks for your reply
>>
>> first, i have about 200,000 users but there is some thin g in the 
>> configuration  that makes the users  connects and disconnects  in 
>> less than 15 minutes , and that makes that huge number of requests
>> (it's a temp. situation ) of-course.
>>
>> second i tried the configuration you sent.
>> the server didn't start, so i guess it OS problem as you said but how 
>> to fix it ??
>> can you help me with that ??
>>
>> thanks again
>> Amr
>>
>> Alan DeKok wrote:
>>> Amr el-Saeed wrote:
>>>  
>>>> Sorry
>>>> the config. was in the first email
>>>>
>>>> I have this configuration
>>>>     
>>> .
>>> ... thread stuff.  There's usually a LOT more configuration than that.
>>>
>>>  
>>>> start_servers = 20
>>>> max_servers = 400
>>>> min_spare_servers = 30
>>>> max_spare_servers = 60
>>>>     
>>>
>>>   I would suggest setting:
>>>
>>>   start_servers = 400
>>>   max_servers = 400
>>>   min_spare_servers = 0
>>>   max_spare_servers = 400
>>>
>>>   If the server doesn't start with those parameters, it's because your
>>> OS doesn't let you start that many threads.  In that case, go fix 
>>> the OS.
>>>
>>>  
>>>> i need 10,000  per second
>>>>     
>>>
>>>   I doubt that very much.  That's a billion packets per day.  Or, it's
>>> 500K users logged in all of the time, each of whom is online for 
>>> only 10
>>> minutes.
>>>
>>>   There are large ISP's with 10+ million users who see only hundreds of
>>> requests per second.  You're talking about 100 times that, which is 
>>> very
>>> unusual.
>>>
>>>   Please explain why you need such a HUGE number of requests.  It may
>>> help solve the problem you're seeing.  Do you have a billion users in
>>> your ISP?  Or do you have fewer users than that?  How many users do you
>>> expect to see online at once?  How many minutes are they online for 
>>> at a
>>> time?
>>>
>>>   Also, most databases can't handle 10k writes per second, (some can't
>>> handle 10k *reads* per second).  So it's very doubtful that you'll be
>>> able to do anything with 10k packets/s, even if the RADIUS server 
>>> itself
>>> can handle them.
>>>
>>>   I'm not aware of a single RADIUS installation on the planet that 
>>> needs
>>> 10K packets per second.  And even the ones that handle hundreds to
>>> thousands of packets per second split those packets among multiple
>>> machines.  i.e. If you have a 10M users in your ISP, you can't 
>>> afford to
>>> have everyone go offline because your ONLY RADIUS server died.  You 
>>> will
>>> need 4-5 RADIUS servers for service stability, at least.
>>>
>>>   The result is that any one RADIUS machine will normally NEVER handle
>>> more than a few hundred packets per second.  If you need more than 
>>> that,
>>> your network is designed wrong.
>>>
>>>   Alan DeKok.
>>> -
>>> List info/subscribe/unsubscribe? See 
>>> http://www.freeradius.org/list/users.html
>>>   
>> -
>> List info/subscribe/unsubscribe? See 
>> http://www.freeradius.org/list/users.html
> -
> List info/subscribe/unsubscribe? See 
> http://www.freeradius.org/list/users.html