Hi,

I have also noticed that under heavy load error will occur more often.
I read somewhere that too small number of MultiPlexer's could make  
problem like this. I have now 20 MultiPlexer's and I haven't seen even  
one sitsuvation when server got more request's then 15 at same time.

> It's quite easy to check if connection is still valid, bu we are unsure
> what to do next in line if we come across such situation. By just hiding
> the error didn't seem to be a fix at a time.

If we know the connection status and we know it's "locked".
Is it possible to send some information to client that it will reconnect?
Same time maybe kill that dead process?

I just increased MaxProcessor's for every <VirtualHost>.
Maybe this make some changes.
At least some of my error's come because of that. I just checked the  
debug output.

I have had this error pretty much at beginning.
Then i was still running 2.0.x apache.
Now this problem is starting to annoying me :)


Quoting Janno Sannik <[EMAIL PROTECTED]>:

> Someone should shed some light on this one, since i have had these
> errors for quite some time.
> Errors seem to occur more under heavier load or on some periods. Some
> days go by with just a 10-15 of those errors and sometimes i get away
> with 40-50 of thouse or more.
> It doesn't seem to be influenced by keepalive patches, since i'm running
> my servers for 19'th day now. Only graceful restarts, no hard restarts
> and situation is unchanged.
> The problem seems to occur then nagle is being disabled: The connection
> seems to be lost (aborted connection?), by the time nagle is being
> disabled and thus the error.
>
>
> Also if server limits are being hit often (low maxprocessors, low
> serverlimit) the errors seem to occur more often.
> It's quite easy to check if connection is still valid, bu we are unsure
> what to do next in line if we come across such situation. By just hiding
> the error didn't seem to be a fix at a time.
>
> Sean (and anyone else running proxy between peruser): Do I understand
> correctly that you don't get these kind of errors?
>
> Just my thoughts.
>
>
> [EMAIL PROTECTED] wrote:
>> Hi,
>>
>> First I have to thank you all for this great MPM.
>> I have been using it now about one year.
>>
>> My problem is error message "Bad file descriptor: apr_socket_opt_set:
>> (TCP_NODELAY)" what appears to ErrorLog. There are about 100-200
>> error's per day.
>> It seems that enough those error will hang apache.
>> Server is on production and running about 150 sites with 7-8Gb   
>> transfer / day.
>> So server is not that busy.
>>
>> I have read all relevant information about this error and try to track
>> it down in apache source. I just can't understunt what is causing this
>> problem.
>> I just turned debug on in peruser source and compiled it again.
>>
>> Im running this system with,
>>
>> - Apache 2.2.6
>> - Peruser 0.3.0rc2
>> - KeepAlive On
>>
>> The problem occurs also with older release 0.2.2 and KeepAlive Off.
>> Actually it seems that this new release make thing even worse.
>> I can post more information about error when I see it next time.
>> I will also post relevant debug messages.
>>
>> I see also about ten to fifteen processor with status UNKNOWN in
>> scoreboard, what is those?
>>
>>
>>
>> _______________________________________________
>> Peruser mailing list
>> [email protected]
>> http://www.telana.com/mailman/listinfo/peruser
>>
>
> _______________________________________________
> Peruser mailing list
> [email protected]
> http://www.telana.com/mailman/listinfo/peruser
>



_______________________________________________
Peruser mailing list
[email protected]
http://www.telana.com/mailman/listinfo/peruser

Reply via email to