I downloaded and installed the latest recommended and security patch 
cluster this morning. Didn't help.

I'm running Solaris 10 Sparc 64-bit

Duane


David L Kensiski wrote:[...]
>>>
>>> The address MUST be resolvable. Try nslookup 127.0.0.1 at the
shell
>>> prompt to check on that.
>>>[...][...]

Thanks guys for quick response.

I was not very precise in what I said before regarding sticky session.
Generally I assume that my Load Balancer )pound) will handle session among a 
few application server.
Let's see an example. User A is redirected to App Server A. The session is 
sticky because of cookies. User A closes the browser and logs back to the 
server 10 minutes later. Will user A be connetced to tha same App Server ? 
Or rather will by handled to another one because the cookies has been 
deleted or expired?. I would like avoid the situation where user is AWLAYS 
redirected to the same App server when new session is established. The goal 
is that subsequnts requestes within session goes tro the same App server but 
each new session shall go to another one.

And second new item. Let's assume I have 4 application server in the back 
end. Can I install the pound on one of them and serve the traffic among 
those 4 servers?

Hopefully that makes my assuption clear.
regards
Mariusz

-----Original Message-----

From: Dave Steinberg <dave(at)redterror.net>

To: pound(at)apsis.ch

Date: Mon, 31 Mar 2008 17:56:31 -0400

Subject: Re: [Pound Mailing List] pounds - few questions - stickiness



[...]
[...]
to 
[...]
[...]
[...]
[...]
[...]
[...]
[...]



Right, right.


[...]
out 
[...]



The backend / your application.


[...]
browser 
[...]
[...]



I believe you're correct, the same cookie will land you on the same 

backend.  That's of course the point of sticky sessions.  :)


[...]
from 
[...]



By priority, or if priorities are equal, I believe its random.


[...]
[...]



poundctl can give you some info, but you'd have to convert to HTML yourself.


[...]
http://www.apsis.ch/pound/index_html
[http://www.apsis.ch/pound/index_html]?



Take a look at the man page included with the source code.  Its quite 

complete.



Regards,
[...]

After a user closes the browser, it depends on how you set your cookies
as to whether they're still there or not.  I think the default is for
cookies to expire at the end of a session (maybe not), but you can set
cookies to expire on a certain date instead (like current_date+1).  If
that's how you set your cookies, I would imagine that if they log back
in to your site, that they'd go back to the original server.  

Just check on how your applications are setting the cookies expiration
time.

Also, I wouldn't recommend running a web server and the load balancer on
the same machine.  I think it can be done if they're running on
different ports, but I still wouldn't recommend it...then again, I
haven't tried it.  If you do try it, I would give the load
balancer/server a low priority since it's doing double duty.

--Alfonso

-----Original Message-----
From: Mariusz Szyszka [mailto:szycha(at)bydnet.com.pl] 
Sent: Tuesday, April 01, 2008 2:01 AM
To: pound(at)apsis.ch; Dave Steinberg
Subject: Re: [Pound Mailing List] pounds - few questions - stickiness

Thanks guys for quick response.

I was not very precise in what I said before regarding sticky session.
Generally I assume that my Load Balancer )pound) will handle session
among a few application server.
Let's see an example. User A is redirected to App Server A. The session
is sticky because of cookies. User A closes the browser and logs back to
the server 10 minutes later. Will user A be connetced to tha same App
Server ? 
Or rather will by handled to another one because the cookies has been
deleted or expired?. I would like avoid the situation where user is
AWLAYS redirected to the same App server when new session is
established. The goal is that subsequnts requestes within session goes
tro the same App server but each new session shall go to another one.

And second new item. Let's assume I have 4 application server in the
back end. Can I install the pound on one of them and serve the traffic
among those 4 servers?

Hopefully that makes my assuption clear.
regards
Mariusz

-----Original Message-----

From: Dave Steinberg <dave(at)redterror.net>

To: pound(at)apsis.ch

Date: Mon, 31 Mar 2008 17:56:31 -0400

Subject: Re: [Pound Mailing List] pounds - few questions - stickiness



[...]
[...]
to 
[...]
[...]
[...]
[...]
[...]
[...]
[...]



Right, right.


[...]
behaviour 
out 
[...]



The backend / your application.


[...]
browser 
[...]
[...]
server. 



I believe you're correct, the same cookie will land you on the same 

backend.  That's of course the point of sticky sessions.  :)


[...]

from 
[...]



By priority, or if priorities are equal, I believe its random.


[...]
[...]



poundctl can give you some info, but you'd have to convert to HTML
yourself.


[...]
http://www.apsis.ch/pound/index_html
[http://www.apsis.ch/pound/index_html]?



Take a look at the man page included with the source code.  Its quite 

complete.



Regards,
[...]

Hmmm... then it's something else.  Well, thanks for trying.  If I get  
a chance, I'll poke at my debugging version again and see if I can  
figure out what's going on.  If you make any progress, please let me  
know.

--Dave


On Mar 31, 2008, at 7:30 PM, Duane Davis wrote:
[...][...]
>>>>
>>>> The address MUST be resolvable. Try nslookup 127.0.0.1 at the
shell
>>>> prompt to check on that.
>>>>
>>>
>>> Doesn't make any difference.
>>>
>>> ] nslookup 127.0.0.1
>>> Server:         10.1.40.1
>>> Address:        10.1.40.1#53
>>>
>>> 1.0.0.127.in-addr.arpa  name = localhost.
>>>
>>>
>>>
>>> ] nslookup localhost
>>> Server:         10.1.40.1
>>> Address:        10.1.40.1#53
>>>
>>> Name:   localhost
>>> Address: 127.0.0.1
>>>
>>>
>>>
>>> ] pound   starting...
>>> line 11: Unknown Listener address "127.0.0.1"
>>>
>>>
>>>
>>>
>>> -- 
>>> To unsubscribe send an email with subject unsubscribe to  
>>> pound(at)apsis.ch.
>>> Please contact roseg(at)apsis.ch for questions.[...][...]

Alfonso, Thank you very much indeed.
New question came up .. it seems to me that I cannot use Pound. I did some 
tests and pound works with http however I was trying load balancer with web 
services in the back end which hosted web page as swf. So the transition 
goes over xml/http.
Is it true that pound does not support LB for swf we pages ?

thanks
Mariusz


-----Original Message-----

From: "Alfonso Espitia" <aespitia(at)castleworldwide.com>

To: <pound(at)apsis.ch>, "Dave Steinberg" <dave(at)redterror.net>

Date: Tue, 1 Apr 2008 07:42:16 -0500

Subject: RE: [Pound Mailing List] pounds - few questions - stickiness




After a user closes the browser, it depends on how you set your cookies

as to whether they're still there or not.  I think the default is for

cookies to expire at the end of a session (maybe not), but you can set

cookies to expire on a certain date instead (like current_date+1).  If

that's how you set your cookies, I would imagine that if they log back

in to your site, that they'd go back to the original server.  



Just check on how your applications are setting the cookies expiration

time.



Also, I wouldn't recommend running a web server and the load balancer on

the same machine.  I think it can be done if they're running on

different ports, but I still wouldn't recommend it...then again, I

haven't tried it.  If you do try it, I would give the load

balancer/server a low priority since it's doing double duty.



--Alfonso



-----Original Message-----

From: Mariusz Szyszka [mailto:szycha(at)bydnet.com.pl] 

Sent: Tuesday, April 01, 2008 2:01 AM

To: pound(at)apsis.ch; Dave Steinberg

Subject: Re: [Pound Mailing List] pounds - few questions - stickiness



Thanks guys for quick response.



I was not very precise in what I said before regarding sticky session.

Generally I assume that my Load Balancer )pound) will handle session

among a few application server.

Let's see an example. User A is redirected to App Server A. The session

is sticky because of cookies. User A closes the browser and logs back to

the server 10 minutes later. Will user A be connetced to tha same App

Server ? 

Or rather will by handled to another one because the cookies has been

deleted or expired?. I would like avoid the situation where user is

AWLAYS redirected to the same App server when new session is

established. The goal is that subsequnts requestes within session goes

tro the same App server but each new session shall go to another one.



And second new item. Let's assume I have 4 application server in the

back end. Can I install the pound on one of them and serve the traffic

among those 4 servers?



Hopefully that makes my assuption clear.

regards

Mariusz



-----Original Message-----



From: Dave Steinberg <dave(at)redterror.net>



To: pound(at)apsis.ch



Date: Mon, 31 Mar 2008 17:56:31 -0400



Subject: Re: [Pound Mailing List] pounds - few questions - stickiness








[...]
[...]


[...]
[...]

to 


[...]
[...]


[...]


[...]


[...]


[...]


[...]


[...]







Right, right.






[...]

behaviour 

out 


[...]







The backend / your application.






[...]

browser 


[...]


[...]

server. 







I believe you're correct, the same cookie will land you on the same 



backend.  That's of course the point of sticky sessions.  :)






[...]



from 


[...]







By priority, or if priorities are equal, I believe its random.






[...]


[...]







poundctl can give you some info, but you'd have to convert to HTML

yourself.






[...]

http://www.apsis.ch/pound/index_html
[http://www.apsis.ch/pound/index_html]

[http://www.apsis.ch/pound/index_html

[http://www.apsis.ch/pound/index_html]]?







Take a look at the man page included with the source code.  Its quite 



complete.







Regards,


[...]

Mariusz Szyszka wrote:[...]

That's not true, pound doesn't care about the content-type of the http 
traffic.

PS - Please keep replies just to the list, no need to CC me personally.

Regards,[...]

I haven't had any trouble using Flash on any websites.  Can you get to
the SWF files if you're NOT going through the load balancer (by directly
hitting the backend server from an internal network)?  
 
For example, if you're using IIS6 (and probably 7), you have to
specifically allow certain extension and mime-types in the server
configuration to allow them to even be served from the server.  My guess
is that you're not using IIS though since you were trying to run pound
on the same machine as one of your servers.
 
Also, I haven't had any issues using XMLHTTP/AJAX through Pound.  
 
One more thing, I don't know if this applies to you or not, but it
appears that the length limit on cookies DECREASED between Pound 1.x and
2.x.  I had to increase this size in the source code (thanks to someone
else on this mailing list) and recompile...everything seems to work for
me now.
 
--Alfonso
[...]

On Tue, 2008-04-01 at 00:52 -0500, Scott Royston wrote:[...]

Yes, it is, and it should be applied to the other session types as well.

Many thanks, this goes into the next release.[...]

Duane,

I finally got around to revisiting this issue.  The problem appears  
that Solaris 10 requires that you provide "hints" when calling the  
getaddrinfo() library call.  Below is a patch to apply to svc.c that  
makes it work under Solaris.  Save this text to a svc.c-patch, then  
run the command "patch -f svc.c-patch svc.c", then make and you  
should be good to go!

Robert, it would be great if you could incorporate this into the next  
release.  It *should* work fine under any OS, but could probably be  
broken out into an OS specific #if.

--Dave


Here's the patch:
*** svc.c-orig  Mon Feb 11 03:53:52 2008
--- svc.c       Mon Apr 14 11:27:26 2008
***************
*** 615,623 ****
   {
       struct addrinfo *chain, *ap;
       int             ret_val;

   #ifdef  HAVE_INET_NTOP
!     if((ret_val = getaddrinfo(name, NULL, NULL, &chain)) == 0) {
           for(ap = chain; ap != NULL; ap = ap->ai_next)
               if(ap->ai_socktype == SOCK_STREAM)
                   break;
--- 615,629 ----
   {
       struct addrinfo *chain, *ap;
       int             ret_val;
+     struct addrinfo hints;

   #ifdef  HAVE_INET_NTOP
!     memset (&hints, 0, sizeof (hints));
!     hints.ai_family = PF_UNSPEC;
!     hints.ai_socktype = SOCK_STREAM;
!     hints.ai_flags |= AI_CANONNAME;
!
!     if((ret_val = getaddrinfo(name, NULL, &hints, &chain)) == 0) {
           for(ap = chain; ap != NULL; ap = ap->ai_next)
               if(ap->ai_socktype == SOCK_STREAM)
                   break;





On Mar 31, 2008, at 7:30 PM, Duane Davis wrote:
[...][...]
>>>>
>>>> The address MUST be resolvable. Try nslookup 127.0.0.1 at the
shell
>>>> prompt to check on that.
>>>>
>>>
>>> Doesn't make any difference.
>>>
>>> ] nslookup 127.0.0.1
>>> Server:         10.1.40.1
>>> Address:        10.1.40.1#53
>>>
>>> 1.0.0.127.in-addr.arpa  name = localhost.
>>>
>>>
>>>
>>> ] nslookup localhost
>>> Server:         10.1.40.1
>>> Address:        10.1.40.1#53
>>>
>>> Name:   localhost
>>> Address: 127.0.0.1
>>>
>>>
>>>
>>> ] pound   starting...
>>> line 11: Unknown Listener address "127.0.0.1"
>>>
>>>
>>>
>>>
>>> -- 
>>> To unsubscribe send an email with subject unsubscribe to  
>>> pound(at)apsis.ch.
>>> Please contact roseg(at)apsis.ch for questions.[...][...]

This patch worked perfectly on my Sun10 install.  I had started making  
the same patch and got side tracked.

Thanks for putting this together Duane.

--Steve

On Apr 14, 2008, at 3:05 PM, David L Kensiski wrote:
[...][...]
>>> I think this is a bug in the Solaris resolver library.  I ran into
 
>>> the exact same problem and put debugging prints into the code and 

>>> the gethostaddr() system call wasn't behaving as I think it  
>>> should.  I am waiting for an outage window to patch my system and 

>>> try again.  If you have the opportunity to patch, I'd love to hear
 
>>> your results.
>>>
>>> --Dave
>>>
>>> On Mar 30, 2008, at 9:04 AM, Duane Davis wrote:
>>>
>>>> Robert Segall wrote:
>>>>>
>>>>> The address MUST be resolvable. Try nslookup 127.0.0.1 at
the  
>>>>> shell
>>>>> prompt to check on that.
>>>>>
>>>>
>>>> Doesn't make any difference.
>>>>
>>>> ] nslookup 127.0.0.1
>>>> Server:         10.1.40.1
>>>> Address:        10.1.40.1#53
>>>>
>>>> 1.0.0.127.in-addr.arpa  name = localhost.
>>>>
>>>>
>>>>
>>>> ] nslookup localhost
>>>> Server:         10.1.40.1
>>>> Address:        10.1.40.1#53
>>>>
>>>> Name:   localhost
>>>> Address: 127.0.0.1
>>>>
>>>>
>>>>
>>>> ] pound   starting...
>>>> line 11: Unknown Listener address "127.0.0.1"
>>>>
>>>>
>>>>
>>>>
>>>> -- 
>>>> 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.[...][...]