This is to announce the release of Pound v2.1.7. Changes in this version
(since 2.1.6):


Enhancements:
- added TCP_NODELAY for faster response times
- added compile-time flag to enable/disable dynamic priorities rescaling

Bug fixes:
- fixed problems with autoconf on some systems
- fixed error in control function (be instead of svc)

The software is at version 2.1.7 (beta quality). Further testing 
(especially under heavy loads), improvements and suggestions are
welcome.
[...]
2.2. In the absence of major problems we'll release 2.2 before the end
of the year.[...]

> Enhancements:[...]

SOL_TCP isn't defined on *BSD systems, as I discovered building on OpenBSD.

 From some quick Google'ing, I came across proftpd's compatability 
stuff, which uses IPPROTO_TCP where SOL_TCP isn't defined:

http://www.castaglia.org/proftpd/doc/devel-guide/src/src/inet.c.html

Changing the SOL_TCP to IPPROTO_TCP yielded a functioning binary.

PS - Just to confirm, is disabling dynamic priority scaling recommended 
or has the "all-requests-end-up-on-one-host" bug been fixed?  In my case 
I'm not yet using priorities.

Regards,[...]

[...]

Was the fixing of the error handling for pthread function calls also 
added to the wish list?
How long would you like to keep the suggested correction in the waiting 
queue of the code maintainers?

Will an overview of open issues become available besides the section 
"other issues" on the page "http://www.apsis.ch/pound/"?
http://en.wikipedia.org/wiki/Comparison_of_ticket-tracking_systems

Regards,
Markus

> This is to announce the release of Pound v2.1.7. Changes in this
version[...]

Thanks for the new release!

The patch below makes my build happy.

Regards,
Simon

--- Pound-2.1.7/Makefile.in.orig        2006-12-06 18:32:18.000000000 +0100
+++ Pound-2.1.7/Makefile.in     2006-12-07 11:02:05.000000000 +0100
(at)(at) -44,7 +44,7 (at)(at)
                ${CC} (at)LDFLAGS(at) -o pound $(OBJS) $(LIBS)

 poundctl:      poundctl.o
-               ${CC} -o poundctl poundctl.o $(LIBS)
+               ${CC} (at)LDFLAGS(at) -o poundctl poundctl.o $(LIBS)

 $(OBJS):       pound.h

On Thu, 2006-12-07 at 11:14 +0100, Simon Matter wrote:[...]

Thanks Simon, this will be fixed in the final release.[...]

On Wed, 2006-12-06 at 13:59 -0500, Dave Steinberg wrote:[...]

Thanks - and we'll also look at using getprotoent() if possible. Next
release to include this.
[...]

We don't recommend disabling the dynamic scaling, but choose according
to your taste. In any case I don't consider it a bug, but rather unusual
behaviour according to circumstances. YMMV: some people seem happy with
it, others less.[...]

Dave Steinberg wrote:[...][...]

Just to follow up, I ended up using the patch below happily on OpenBSD. 
  There may be better ways of doing this more portably, but this is just 
a start.

Regards,[...]

>> Changing the SOL_TCP to IPPROTO_TCP yielded a functioning binary.[...]

Sounds good - I expect that will be more portable than my solution.
[...][...]

Ok, I'll play around and see how it feels.  I suppose I'll have to 
figure out how to get Cacti to do some of those fancy multi-host 
reqs/sec graphs and evaluate.  :)

Regards,[...]

On Thu, 2006-12-07 at 13:05 -0500, Dave Steinberg wrote:[...]

I hope you'll share with us all your results - it would be really
helpful.[...]

This is to announce the release of Pound v2.1.8. This is primarily a
bug-fix release. Changes in this version (since 2.1.7):


Bug fixes:
- fixed problems with autoconf on some systems (LDFLAGS)
- fixed the problem of undefined SOL_TCP on BSD/Solaris

The software is at version 2.1.8 (pre-release quality). Further testing 
(especially under heavy loads), improvements and suggestions are
welcome.

We consider this to be a second release candidate for 2.2. In the
absence of major problems we'll release 2.2 before the end
of the year.[...]

This is to announce the release of Pound v2.2. Changes in this version
(since 2.1.8):

Enhancements:
  - added the host to LogLevel 2 (if available)
  - added support for tcmalloc (from the Google perftools package)

Bug fixes:
  - fixed problem with the initialisation of host_mut

*************************************
Cumulative changes since version 2.1:
*************************************

Enhancements:
  - added dynamic rescaling of back-end priorities, compile-time flag to
    enable/disable it
  - added support for emergency back-ends
  - the program poundctl(8) is now available, added the Control
    configuration directive
  - SESS_IP now behaves like other session types (no longer sticky)
  - added RewriteLocation 2: rewrite location if it points to same
    address, butignore port
  - Redirect uses the original request path
  - added RewriteDestination configuration flag to enable rewriting the
    Destination header
  - removed msdav compile-time configuration flag and MSDAV
    configuration flag,extended xHTTP to allow for WebDAV, MS-DAV and
    MS-RPC
  - added CRLlist directive, split CRL from CA
  - Error replies are sent as pure HTML
  - split error messages into:
    - LOG_ERR: errors (mostly fatal)
    - LOG_WARNING: errors (non-fatal)
    - LOG_NOTICE: problems
    - LOG_INFO: extra information
  - time to serve the requests is logged in LogLevel 2
  - added the (virtual) host to LogLevel 2 (if available)
  - added line numbers to config error messages
  - added TCP_NODELAY for faster response times
  - added support for tcmalloc (from the Google perftools package)

Bug fixes:
  - fixed problem in str_be (evident mostly in LogLevel 2)
  - added 'const' wherever necessary
  - check for errors in mutex handling
  - fixed the verb pattern in HTTPS listeners
  - content is now ignored only on HEAD requests
  - fixed problems with autoconf on some systems
  - fixed problem with the initialisation of host_mut

The software is at version 2.2 (production quality). Further testing
(especially under heavy loads), improvements and suggestions are
welcome.[...]

On 16-dec-2006, at 10:58, Robert Segall wrote:
[...]
I just finished building the rpms for Fedora 5 and 6, they should be  
available on the fedora mirrors in a few hours.
$ sudo yum update pound will update to 2.2

Cheers,

Ruben Kerkhof

Robert Segall wrote:[...]

there is still a minor warning when compiling using Sun C:
"poundctl.c", line 41: warning: old-style declaration or incorrect type 
for: main

Hi Robert,

First many thanks for your work on this excellent software.

A little while ago I reported a problem with poundctl and OpenBSD, that 
being that when I specify the 'Control' directive in my Pound config I 
am no longer able to connect to the backend servers. Poundctl 
successfully connects to the running Pound instance through the socket, 
reporting backend servers and even reporting the client to backend 
mapping! however the client never connects and simply times out.

'file' reports the file created as a 'socket' so that looks ok. Removing 
the 'control' line from my config everything springs back to life so it 
looks like either I'm not doing something I need to with the socket file 
or that there is a problem with this on OpenBSD.

Any suggestions as the where I should look next? Did you manage build 
your OpenBSD 4.0 box and test Pound?

Thanks again.

Simon



Robert Segall wrote:[...]

On Thu, 2006-12-28 at 12:20 +0000, Simon Slaytor wrote:[...]

Yes, we did, no, we had no problems: Pound 2.2 compiled out-of-the-box
and ran without any problems, including with the control socket
activated. The requests were passed just fine and poundctl worked as
expected.

I suggest you have a look at your pf rules - maybe they block your
access to the back-end?

Could anyone else who uses Pound on OpenBSD confirm this?[...]

Hi Robert,

Thanks for the reply, the pound box is internal so no PF at all to 
complicate matters.

Yes no problems building and installing just in using poundctl.

Do I have to do any 'magic' on the socket file? Initially I don't create 
the file, instead letting Pound create it when first run is this correct?

The OBSD box is a vanilla install with Pound so nothing should be 
different from our two installs.

Regards

Simon




Robert Segall wrote:[...][...][...]

On Fri, 2006-12-29 at 09:36 +0000, Simon Slaytor wrote:[...]

Yes, Pound will create the socket by itself - you don't need to do
anything.
[...]

Puzzling indeed. The most puzzling thing is that your poundctl shows
active connections - which means Pound actually contacted your back-end.
I can't imagine any reason for it not to return the reply to the client.
Did you try telnet directly to the Pound listener just to make sure
there really is no response?[...]