|
/
Zope
/
Apsis
/
Pound Mailing List
/
Archive
/
2011
/
2011-11
/
TimeOut apparently not working in 2.5 and 2.6f
[
we have a strange loop / Pat Erler ... ]
[
Directory with headrequire / Adrian Padilla ... ]
TimeOut apparently not working in 2.5 and 2.6f
Dominik Klein <dominik.klein(at)googlemail.com> |
2011-11-22 12:27:58 |
[ SNIP ]
|
Hi
I am evaluating pound for a project and while playing with the TimeOut
option, I found that it does not seem to work.
No matter what I set it to and no matter whether I set it globally or
specifically for a BackEnd Server, the used TimeOut always seems to be
15 seconds.
My Test Setup is
/> Apache 1
Client -> Pound -> Apache 2
\> Apache 3
On Apache 1 I configure iptables to DROP incoming tcp port 80.
First of all I noticed that pound does not realize that the BackEnd is
dead until the first request is tried towards this BackEnd. Maybe that's
by design, but in case it's not I figured I'd mention what I saw.
Then, once the Client puts a request that pound decides to forward to
Apache 1, it takes 15.00x seconds for the request to complete, no matter
what or in which place the TimeOut is configured. I tested with 180 and
10 seconds.
Logs look like this:
(41f47950) connect_nb: poll timed out
(41f47950) backend 10.250.250.198:80 connect: Connection timed out
(41f47950) BackEnd 10.250.250.198:80 dead (killed)
10.250.250.191 10.250.250.235 - - [22/Nov/2011:12:05:54 +0100] "GET /
HTTP/1.1" 200 66 "" "Links (2.1pre37; Linux 2.6.26-2-amd64 x86_64;
80x22)" (- -> 10.250.250.221:80) 15.003 sec
This test was done with version 2.5 and 2.6f and both show the same
behaviour. Configuration file is:
User "www-data"
Group "www-data"
LogLevel 5
Daemon 0
LogFacility -
Alive 2
Client 30
TimeOut 180
# TimeOut 10
DynScale 1
ListenHTTP
Address 10.250.250.191
Port 80
HeadRemove "^Host:"
addHeader "host: test.mydomain.de"
Service
Session
Type IP
TTL 6000
End
BackEnd
Address 10.250.250.165
Port 80
End
BackEnd
Address 10.250.250.198
Port 80
TimeOut 180
# TimeOut 10
End
BackEnd
Address 10.250.250.221
Port 80
End
End
End
Did I do something wrong or is this a bug? If it is a bug: Is there a
bug tracker for pound that I could post this to?
Regards
Dominik
|
|
|
RE: [Pound Mailing List] TimeOut apparently not working in 2.5 and 2.6f
Joe Gooch <mrwizard(at)k12system.com> |
2011-11-22 15:12:01 |
[ SNIP ]
|
Try setting the ConnTO option explicitly.
Joe
> -----Original Message-----
> From: Dominik Klein [mailto:dominik.klein(at)googlemail.com]
> Sent: Tuesday, November 22, 2011 6:28 AM
> To: pound(at)apsis.ch
> Subject: [Pound Mailing List] TimeOut apparently not working in 2.5 and
> 2.6f
>
> Hi
>
> I am evaluating pound for a project and while playing with the TimeOut
> option, I found that it does not seem to work.
>
> No matter what I set it to and no matter whether I set it globally or
> specifically for a BackEnd Server, the used TimeOut always seems to be
> 15 seconds.
>
> My Test Setup is
>
> /> Apache 1
> Client -> Pound -> Apache 2
> \> Apache 3
>
> On Apache 1 I configure iptables to DROP incoming tcp port 80.
>
> First of all I noticed that pound does not realize that the BackEnd is
> dead until the first request is tried towards this BackEnd. Maybe
> that's
> by design, but in case it's not I figured I'd mention what I saw.
>
> Then, once the Client puts a request that pound decides to forward to
> Apache 1, it takes 15.00x seconds for the request to complete, no
> matter
> what or in which place the TimeOut is configured. I tested with 180 and
> 10 seconds.
>
> Logs look like this:
>
> (41f47950) connect_nb: poll timed out
> (41f47950) backend 10.250.250.198:80 connect: Connection timed out
> (41f47950) BackEnd 10.250.250.198:80 dead (killed)
> 10.250.250.191 10.250.250.235 - - [22/Nov/2011:12:05:54 +0100] "GET /
> HTTP/1.1" 200 66 "" "Links (2.1pre37; Linux 2.6.26-2-amd64 x86_64;
> 80x22)" (- -> 10.250.250.221:80) 15.003 sec
>
> This test was done with version 2.5 and 2.6f and both show the same
> behaviour. Configuration file is:
>
> User "www-data"
> Group "www-data"
> LogLevel 5
> Daemon 0
> LogFacility -
> Alive 2
> Client 30
> TimeOut 180
> # TimeOut 10
> DynScale 1
> ListenHTTP
> Address 10.250.250.191
> Port 80
> HeadRemove "^Host:"
> addHeader "host: test.mydomain.de"
> Service
> Session
> Type IP
> TTL 6000
> End
> BackEnd
> Address 10.250.250.165
> Port 80
> End
> BackEnd
> Address 10.250.250.198
> Port 80
> TimeOut 180
> # TimeOut 10
> End
> BackEnd
> Address 10.250.250.221
> Port 80
> End
> End
> End
>
> Did I do something wrong or is this a bug? If it is a bug: Is there a
> bug tracker for pound that I could post this to?
>
> Regards
> Dominik
>
> --
> To unsubscribe send an email with subject unsubscribe to
> pound(at)apsis.ch.
> Please contact roseg(at)apsis.ch for questions.
|
|
|
Re: [Pound Mailing List] TimeOut apparently not working in 2.5 and 2.6f
Dominik Klein <dominik.klein(at)googlemail.com> |
2011-11-23 07:15:55 |
[ SNIP ]
|
On 11/22/2011 03:12 PM, Joe Gooch wrote:
> Try setting the ConnTO option explicitly.
This works for both my testcases (10 and 180).
Thanks
Dominik
> Joe
>
>
>> -----Original Message-----
>> From: Dominik Klein [mailto:dominik.klein(at)googlemail.com]
>> Sent: Tuesday, November 22, 2011 6:28 AM
>> To: pound(at)apsis.ch
>> Subject: [Pound Mailing List] TimeOut apparently not working in 2.5 and
>> 2.6f
>>
>> Hi
>>
>> I am evaluating pound for a project and while playing with the TimeOut
>> option, I found that it does not seem to work.
>>
>> No matter what I set it to and no matter whether I set it globally or
>> specifically for a BackEnd Server, the used TimeOut always seems to be
>> 15 seconds.
>>
>> My Test Setup is
>>
>> /> Apache 1
>> Client -> Pound -> Apache 2
>> \> Apache 3
>>
>> On Apache 1 I configure iptables to DROP incoming tcp port 80.
>>
>> First of all I noticed that pound does not realize that the BackEnd is
>> dead until the first request is tried towards this BackEnd. Maybe
>> that's
>> by design, but in case it's not I figured I'd mention what I saw.
>>
>> Then, once the Client puts a request that pound decides to forward to
>> Apache 1, it takes 15.00x seconds for the request to complete, no
>> matter
>> what or in which place the TimeOut is configured. I tested with 180 and
>> 10 seconds.
>>
>> Logs look like this:
>>
>> (41f47950) connect_nb: poll timed out
>> (41f47950) backend 10.250.250.198:80 connect: Connection timed out
>> (41f47950) BackEnd 10.250.250.198:80 dead (killed)
>> 10.250.250.191 10.250.250.235 - - [22/Nov/2011:12:05:54 +0100] "GET /
>> HTTP/1.1" 200 66 "" "Links (2.1pre37; Linux 2.6.26-2-amd64 x86_64;
>> 80x22)" (- -> 10.250.250.221:80) 15.003 sec
>>
>> This test was done with version 2.5 and 2.6f and both show the same
>> behaviour. Configuration file is:
>>
>> User "www-data"
>> Group "www-data"
>> LogLevel 5
>> Daemon 0
>> LogFacility -
>> Alive 2
>> Client 30
>> TimeOut 180
>> # TimeOut 10
>> DynScale 1
>> ListenHTTP
>> Address 10.250.250.191
>> Port 80
>> HeadRemove "^Host:"
>> addHeader "host: test.mydomain.de"
>> Service
>> Session
>> Type IP
>> TTL 6000
>> End
>> BackEnd
>> Address 10.250.250.165
>> Port 80
>> End
>> BackEnd
>> Address 10.250.250.198
>> Port 80
>> TimeOut 180
>> # TimeOut 10
>> End
>> BackEnd
>> Address 10.250.250.221
>> Port 80
>> End
>> End
>> End
>>
>> Did I do something wrong or is this a bug? If it is a bug: Is there a
>> bug tracker for pound that I could post this to?
>>
>> Regards
>> Dominik
>>
>> --
>> To unsubscribe send an email with subject unsubscribe to
>> pound(at)apsis.ch.
>> Please contact roseg(at)apsis.ch for questions.
>
> --
> To unsubscribe send an email with subject unsubscribe to pound(at)apsis.ch.
> Please contact roseg(at)apsis.ch for questions.
|
|
|
RE: [Pound Mailing List] TimeOut apparently not working in 2.5 and 2.6f
Joe Gooch <mrwizard(at)k12system.com> |
2011-11-23 20:04:28 |
[ SNIP ]
|
The man page should probably be updated... It says connto defaults to
timeout... Which when I look at the code... isn't true. They both default to
15. But if you change the timeout the connto won't inherit it.
I would think to remove ambiguity you'd always specify both explicitly in your
configurations.
Joe
> -----Original Message-----
> From: Dominik Klein [mailto:dominik.klein(at)googlemail.com]
> Sent: Wednesday, November 23, 2011 1:16 AM
> To: pound(at)apsis.ch
> Subject: Re: [Pound Mailing List] TimeOut apparently not working in 2.5
> and 2.6f
>
> On 11/22/2011 03:12 PM, Joe Gooch wrote:
> > Try setting the ConnTO option explicitly.
>
> This works for both my testcases (10 and 180).
>
> Thanks
> Dominik
>
> > Joe
> >
> >
> >> -----Original Message-----
> >> From: Dominik Klein [mailto:dominik.klein(at)googlemail.com]
> >> Sent: Tuesday, November 22, 2011 6:28 AM
> >> To: pound(at)apsis.ch
> >> Subject: [Pound Mailing List] TimeOut apparently not working in 2.5
> and
> >> 2.6f
> >>
> >> Hi
> >>
> >> I am evaluating pound for a project and while playing with the
> TimeOut
> >> option, I found that it does not seem to work.
> >>
> >> No matter what I set it to and no matter whether I set it globally
> or
> >> specifically for a BackEnd Server, the used TimeOut always seems to
> be
> >> 15 seconds.
> >>
> >> My Test Setup is
> >>
> >> /> Apache 1
> >> Client -> Pound -> Apache 2
> >> \> Apache 3
> >>
> >> On Apache 1 I configure iptables to DROP incoming tcp port 80.
> >>
> >> First of all I noticed that pound does not realize that the BackEnd
> is
> >> dead until the first request is tried towards this BackEnd. Maybe
> >> that's
> >> by design, but in case it's not I figured I'd mention what I saw.
> >>
> >> Then, once the Client puts a request that pound decides to forward
> to
> >> Apache 1, it takes 15.00x seconds for the request to complete, no
> >> matter
> >> what or in which place the TimeOut is configured. I tested with 180
> and
> >> 10 seconds.
> >>
> >> Logs look like this:
> >>
> >> (41f47950) connect_nb: poll timed out
> >> (41f47950) backend 10.250.250.198:80 connect: Connection timed out
> >> (41f47950) BackEnd 10.250.250.198:80 dead (killed)
> >> 10.250.250.191 10.250.250.235 - - [22/Nov/2011:12:05:54 +0100] "GET
> /
> >> HTTP/1.1" 200 66 "" "Links (2.1pre37; Linux 2.6.26-2-amd64 x86_64;
> >> 80x22)" (- -> 10.250.250.221:80) 15.003 sec
> >>
> >> This test was done with version 2.5 and 2.6f and both show the same
> >> behaviour. Configuration file is:
> >>
> >> User "www-data"
> >> Group "www-data"
> >> LogLevel 5
> >> Daemon 0
> >> LogFacility -
> >> Alive 2
> >> Client 30
> >> TimeOut 180
> >> # TimeOut 10
> >> DynScale 1
> >> ListenHTTP
> >> Address 10.250.250.191
> >> Port 80
> >> HeadRemove "^Host:"
> >> addHeader "host: test.mydomain.de"
> >> Service
> >> Session
> >> Type IP
> >> TTL 6000
> >> End
> >> BackEnd
> >> Address 10.250.250.165
> >> Port 80
> >> End
> >> BackEnd
> >> Address 10.250.250.198
> >> Port 80
> >> TimeOut 180
> >> # TimeOut 10
> >> End
> >> BackEnd
> >> Address 10.250.250.221
> >> Port 80
> >> End
> >> End
> >> End
> >>
> >> Did I do something wrong or is this a bug? If it is a bug: Is there
> a
> >> bug tracker for pound that I could post this to?
> >>
> >> Regards
> >> Dominik
> >>
> >> --
> >> To unsubscribe send an email with subject unsubscribe to
> >> pound(at)apsis.ch.
> >> Please contact roseg(at)apsis.ch for questions.
> >
> > --
> > To unsubscribe send an email with subject unsubscribe to
> pound(at)apsis.ch.
> > Please contact roseg(at)apsis.ch for questions.
>
> --
> To unsubscribe send an email with subject unsubscribe to
> pound(at)apsis.ch.
> Please contact roseg(at)apsis.ch for questions.
|
|
|
|
