I all

I have a pound in front of several apache
sometime apache report error 500 (application problem)

How can I do to consider node in offline if it return errorcode500?


kind rega

It might be timing out if your script takes a long time to run.  

I think the default time is 30 seconds.  If your script takes longer
than 30 seconds to run (without going through pound), then try
increasing the TimeOut period.

Do the same pages work fine if you're NOT going through pound? 

-----Original Message-----
From: Martin [mailto:list(at)manuelmartini.it] 
Sent: Tuesday, July 08, 2008 9:04 AM
To: pound(at)apsis.ch
Subject: [Pound Mailing List] pound and errorcode 500


I all

I have a pound in front of several apache sometime apache report error
500 (application problem)

How can I do to consider node in offline if it return errorcode500?


kind rega

--
To unsubscribe send an email with subject unsubscribe to pound(at)apsis.ch.
Please contact roseg(at)apsis.ch for questions.

--
This message has been scanned for viruses and dangerous content by
SecureMail, and is believed to be clean.

Alfonso Espitia ha scritto:[...]

apache report error 500 everytime application  encure an exception (I 
can't do changes  to apps at this time)

[...]

yes every time app have an error report a 500 error

I would like to consider error 500 like node down


kind regards

------------------------
[...]

yes every time app have an error report a 500 error

------------------------

If I'm understanding this correctly, then your pages are currently NOT
working fine (even without pound).  You, or whoever wrote the
script/application should check into those errors and work through the
exceptions (what's causing your application to crash).  Pound isn't
going to fix an application error.  

Once you get your application working correctly (without pound), I would
then retry it with pound.  If you're using sessions or cookies, I would
also attempt using the session tracking feature in Pound.

--Alfonso



 

-----Original Message-----
From: Martin [mailto:list(at)manuelmartini.it] 
Sent: Tuesday, July 08, 2008 9:37 AM
To: pound(at)apsis.ch
Subject: Re: [Pound Mailing List] pound and errorcode 500

Alfonso Espitia ha scritto:[...]

apache report error 500 everytime application  encure an exception (I
can't do changes  to apps at this time)

[...]

yes every time app have an error report a 500 error

I would like to consider error 500 like node down


kind regards



--
To unsubscribe send an email with subject unsubscribe to pound(at)apsis.ch.
Please contact roseg(at)apsis.ch for questions.
[...]

[...]


He is not asking for pound to fix anything. He is asking for pound to 
consider the backend down (as if it was completely unreachable) when his 
application reports a 500. It is actually a good feature request.

Rick

Rick Blundell ha scritto:[...][...][...]

exatly :) :)

sometime I can't do anythink on apps layer


kind regards

I guess I need more background info.  At the moment I don't see how that
would be a good feature request.

If I have 20 apps running on a machine, and one of them is crashing, I
wouldn't want that server to be considered "offline" since I have 19
other apps that I want load balanced.  I just need to fix the 1.

--Alfonso 

-----Original Message-----
From: Rick Blundell [mailto:rickb(at)rapidvps.com] 
Sent: Tuesday, July 08, 2008 10:33 AM
To: pound(at)apsis.ch
Subject: Re: [Pound Mailing List] pound and errorcode 500

[...]
[...]


He is not asking for pound to fix anything. He is asking for pound to
consider the backend down (as if it was completely unreachable) when his
application reports a 500. It is actually a good feature request.

Rick

--
To unsubscribe send an email with subject unsubscribe to pound(at)apsis.ch.
Please contact roseg(at)apsis.ch for questions.

--
This message has been scanned for viruses and dangerous content by
SecureMail, and is believed to be clean.

[...]

Agreed.  We had this discussion before, using the 5xx HTTP code to 
determine the health of the server is not a very good idea.  Pound has a 
HAPort option, so you can write an app which will check the health of 
your service.  If everything is OK, the app will allow pound to connect 
to it on the HAPort.  If not, then it won't allow connections.  We do 
something similar here. 

You can also use such application to take machines out of the pool, 
without stopping your backend http server.

Alfonso Espitia ha scritto:[...]

I would like to see  that feature as option to set in pound.cfg  :):)

The feature is you can write a script to monitor your servers however you want,
to determine whether they're alive or not.  If that means checking the last 5
minutes of pound logs, that's fine.  If that means using LWP to do a query
against the backend and look for a 500 error, that's fine too.

A 500 error could be limited to one page.  And if it's a complexity issue, it
doesn't matter which one of my backends you'd use... They'd all do the same
thing.  Failing a server for one user's one session's page failing would be a
horrible idea.  Of course, that's my environment.

That's why the life checks are usually based on scripts that are implementation
specific.

-G

[...]

Indeed,

some sort of evaluation of backends health from its HTTP responce would 
not be a key feature in terms of usage frequency, but still irreplacible 
at some cases.

j.

Martin rašė:[...][...][...]

In our environment, we only have one single application on the backend  
servers. If it fails on one server, that instance needs to be taken  
out of the pool. In other words, HTTP response code interpretation is  
most definitively useful.

Just my 2 cents.

/Eirik


On Jul 9, 2008, at 08:31, tyntas wrote:
[...][...]
>>> I guess I need more background info.  At the moment I don't see  
>>> how that
>>> would be a good feature request.
>>>
>>> If I have 20 apps running on a machine, and one of them is  
>>> crashing, I
>>> wouldn't want that server to be considered "offline" since I have
19
>>> other apps that I want load balanced.  I just need to fix the 1.
>>>
>>>[...][...]