Mailing List Archive

Hmm... This isn't in the design of mod_backhand. mod_backhand doesn't require
any connectivity tothe outside world.. Are you sure that this traffic was not
there before you got mod_backhand running?

The only thing I can think of that would cause this inside of mod_backhand is
the MulticastStats line. If you MulticastStats and use *two* arguements, the
first is considered to be the IP address of the local server (in case it can't
figure that out for itself).

So, if you server has two IP address 10.0.0.101 and 192.168.10.101 and it
*thinks* it is 10.0.0.101, but you want it to be using the other
192.168.10.101 interface for backhand redirecting, then you could specify:
MulticastStats 192.168.10.101 192.168.10.255:4445

Then, all of the fellow machines will see this machines as 192.168.10.101
instead of 10.0.0.101.

So, if you had a MulticastStats line where the first arguement was the IP
address of that Akamai server, then the machine would be advcertising itself
as the akamai server. When other machines within the cluster want to redirect
to it will redirect to the akamai server instead.

What do the IP addresses look like in the /backhand/ status page?

Markus Neteler wrote:
> after installing backhand 1.1.0 and latest Apache successfully
> on Linux and SUN/Solaris (they see each other), I found
> obscure and intensive network traffic by using ethereal.
>
> Continuously the server
> a195-206-69-39.deploy.akamaitechnologies.com
> was contacted:
>
> a195-206-69-39.deploy.akamaitechnologies.com rapido-ip -> http [ACK] seq=... acq=...
>
> As I am rather unfamiliar with the details of backhand, this might
> be o.k. But it is not documented anywhere.
>
> Perhaps someone is willing to explain this,

--
Theo Schlossnagle
1024D/A8EBCF8F/13BD 8C08 6BE2 629A 527E 2DC2 72C2 AD05 A8EB CF8F
2047R/33131B65/71 F7 95 64 49 76 5D BA 3D 90 B9 9F BE 27 24 E7

Theo,

thanks for your quick response!

On Wed, Nov 08, 2000 at 03:44:12PM -0500, Theo E. Schlossnagle wrote:
> Hmm... This isn't in the design of mod_backhand. mod_backhand doesn't require
> any connectivity tothe outside world.. Are you sure that this traffic was not
> there before you got mod_backhand running?
unfortunately I tested after installing backhand. But it's good to know
that this is not the intention of backhand :-) Actually the traffic
is gone.

> So, if you had a MulticastStats line where the first arguement was the IP
> address of that Akamai server, then the machine would be advcertising itself
> as the akamai server. When other machines within the cluster want to redirect
> to it will redirect to the akamai server instead.
No, it was a different address. But...

> What do the IP addresses look like in the /backhand/ status page?
..to be honest: I don't fully understand the concept of MulticastStats.
The address to specify, is it one of my servers? Sorry, but I didn't get
it from the docs/FAQ.

The situation:
I have my main server:
130.75.72.14
(SUN)

My backup-server is
130.75.72.36
(Linux, identical copy)

On 130.75.72.14 I have defined in httpd.conf:
[...]

UnixSocketDir /usr/local/apache/backhand
MulticastStats 130.75.72.36:4445,1
AcceptStats 130.75.72.0/80

On 130.75.72.36 I have defined in httpd.conf:
[...]
UnixSocketDir /var/backhand/backhand
MulticastStats 130.75.72.14:4445,1
AcceptStats 130.75.72.0/80

Looking at /backhand/:
http://bermuda.geog.uni-hannover.de/backhand/
0 bermuda.geog.uni-hannover.de 973765382 130.75.72.36:80 200900608 49161408 0/0 0 [0] 501697 1 0.000000/1 0.000000
(color red)

the www (aka hgeo02) is not seen...

Same comparing situation on www.geog.uni-hannover.de

http://www.geog.uni-hannover.de/cgi-bin/backhand-test
-> o.k.

http://bermuda.geog.uni-hannover.de/cgi-bin/backhand-test
-> o.k.

Servers are o.k., too.

Using
MulticastStats 225.220.221.20:4445,1
on both servers backhand is running. But I am afraid that I am not allowed
to use this address... Or is it reserved for such activities?
nslookup tells me: 225.220.221.20: Non-existent host/domain. Or do I
have to calculate a proper address due to my network?

I would be glad if you could tell me about the MulticastStats secret
(and maybe put an live example onto the FAQ).

Thanks in advance

Markus Neteler

PS: Final question: If backhand is running (as it did with the
225.220.221.20 address), how can I test the environment? Run some
CPU-intensive application on the main server? Or generate many web
accesses (how to do that?)? Just curious...

--
Dipl.-Geogr. Markus Neteler * University of Hannover
Institute of Physical Geography and Landscape Ecology
Schneiderberg 50 * D-30167 Hannover * Germany
Tel: ++49-(0)511-762-4494 Fax: -3984

Hi again,

when configuratin both servers to
MulticastStats 225.220.221.20:4445,1

the backhand is running. But am I allowed to use
this address?

"ethereal" is telling me:
bermuda.geog.uni-hannover.de 225.220.221.20 UDP Source port: upnotify Destination port: upnotify
www.geog.uni-hannover.de 225.220.221.20 UDP Source port: upnotify Destination port: upnotify

Is that o.k?

Thank again

Markus

--
Dipl.-Geogr. Markus Neteler * University of Hannover
Institute of Physical Geography and Landscape Ecology
Schneiderberg 50 * D-30167 Hannover * Germany
Tel: ++49-(0)511-762-4494 Fax: -3984

Markus Neteler wrote:
> The situation:
> I have my main server:
> 130.75.72.14
> (SUN)
>
> My backup-server is
> 130.75.72.36
> (Linux, identical copy)
>
> On 130.75.72.14 I have defined in httpd.conf:
> [...]
>
> UnixSocketDir /usr/local/apache/backhand
> MulticastStats 130.75.72.36:4445,1
> AcceptStats 130.75.72.0/80
>
> On 130.75.72.36 I have defined in httpd.conf:
> [...]
> UnixSocketDir /var/backhand/backhand
> MulticastStats 130.75.72.14:4445,1
> AcceptStats 130.75.72.0/80
>
> Looking at /backhand/:
> http://bermuda.geog.uni-hannover.de/backhand/
> 0 bermuda.geog.uni-hannover.de 973765382 130.75.72.36:80 200900608 49161408 0/0 0 [0] 501697 1 0.000000/1 0.000000
> (color red)
>
> the www (aka hgeo02) is not seen...
>
> Same comparing situation on www.geog.uni-hannover.de
> http://www.geog.uni-hannover.de/cgi-bin/backhand-test
> -> o.k.
> http://bermuda.geog.uni-hannover.de/cgi-bin/backhand-test
> -> o.k.

This is fine.

> Using
> MulticastStats 225.220.221.20:4445,1
> on both servers backhand is running. But I am afraid that I am not allowed
> to use this address... Or is it reserved for such activities?
> nslookup tells me: 225.220.221.20: Non-existent host/domain. Or do I
> have to calculate a proper address due to my network?
>
> I would be glad if you could tell me about the MulticastStats secret
> (and maybe put an live example onto the FAQ).

This is fine to.. IP's in the range: (224.0.0.0- 239.255.255.255) are reserved
for IP mulicast. They are routed differently and are not *allowed* to be
assigned to a machine. So you can use 225.220.221.20 without much problem.
The other option is to use your broadcast address.

The only time you need to specify *two* arguments to MulticastStats is when
you have more than one IP address assigned to a machine and mod_backhand
"chooses" the wrong one. I don't think you'll need to bother with this right
now.

> PS: Final question: If backhand is running (as it did with the
> 225.220.221.20 address), how can I test the environment? Run some
> CPU-intensive application on the main server? Or generate many web
> accesses (how to do that?)? Just curious...

One easy way to tell if it is wokring is to use the byLoad candidacy function
and (assuming the servers are both unloaded) run "perl -e 'while(1) {;}'" on
one server. This will drive the load up on that machine. Make some requests
to that machine and they should be services by the other machine. There is a
backhand-test cgi that comes with mod_backhand... It will tell you where you
entered the cluster and how eventually serviced your request.

Try httperf or apache's "ab".

Hmmm, I am working on a project called daiquiri that will re run Apache common
logs against a cluster of web servers (or just one) and you can "multiply" the
log times a factor (make every request twice or a hundred times)... It is
available out of CVS at :pserver:anonymous@commedia.cnds.jhu.edu:/storage/cvs
and the module is called daiquiri. If you don't know how to use CVS, I am
afraid that you will have to wait for the first public release which isn't yet
scheduled.

--
Theo Schlossnagle
1024D/A8EBCF8F/13BD 8C08 6BE2 629A 527E 2DC2 72C2 AD05 A8EB CF8F
2047R/33131B65/71 F7 95 64 49 76 5D BA 3D 90 B9 9F BE 27 24 E7

Markus Neteler wrote:
> when configuratin both servers to
> MulticastStats 225.220.221.20:4445,1
>
> the backhand is running. But am I allowed to use
> this address?

Yes, you can use any address in the assigned IP Multicast block
224.0.0.0->239.255.255.255

You can also use your broadcast address (leave the ,1 off the end though).
Multicast is often more effecient than broadcast. That is why the option
exists.

> "ethereal" is telling me:
> bermuda.geog.uni-hannover.de 225.220.221.20 UDP Source port: upnotify Destination port: upnotify
> www.geog.uni-hannover.de 225.220.221.20 UDP Source port: upnotify Destination port: upnotify

Apparently 225.220.221.20 is assigned to upnotify... Doesn't bother me :-)

--
Theo Schlossnagle
1024D/A8EBCF8F/13BD 8C08 6BE2 629A 527E 2DC2 72C2 AD05 A8EB CF8F
2047R/33131B65/71 F7 95 64 49 76 5D BA 3D 90 B9 9F BE 27 24 E7

Quoting Markus Neteler (Sat, Nov 11, 2000 at 01:53:09PM +0000)

> http://www.geog.uni-hannover.de/backhand/
> says: #ready/#used servers: 0/0
> Although the machine is running properly. Mhhh.

apachectl graceful will "fix" that

Colin.

--

Colin May - rael@wibble.org.uk - http://www.wibble.org.uk
"Look across at dreamland and tell me what you see..."

Theo,

thanks for your useful explanations! It seems the obscure traffic
was there for any other reason (mhh, better I don't knwo why).


On Thu, Nov 09, 2000 at 01:12:53PM -0500, Theo E. Schlossnagle wrote:
[...]
> > PS: Final question: If backhand is running (as it did with the
> > 225.220.221.20 address), how can I test the environment? Run some
> > CPU-intensive application on the main server? Or generate many web
> > accesses (how to do that?)? Just curious...
>
> One easy way to tell if it is wokring is to use the byLoad candidacy function
> and (assuming the servers are both unloaded) run "perl -e 'while(1) {;}'" on
> one server. This will drive the load up on that machine. Make some requests
> to that machine and they should be services by the other machine. There is a
> backhand-test cgi that comes with mod_backhand... It will tell you where you
> entered the cluster and how eventually serviced your request.
Yes, I have seen that script.

> Try httperf or apache's "ab".

To test performance I tried:
httperf --server=www.geog.uni-hannover.de --num-conns=1000 --num-calls=1000

Still the cgi-script tells me that www would be serving my request.
I am unable to drive the load such high to get backhand moving my requests
to the other machine :-)

Funny thing is that:
http://www.geog.uni-hannover.de/backhand/

says: #ready/#used servers: 0/0

Although the machine is running properly. Mhhh.

Sorry to bother you here again, perhaps you have an idea for that.

> Hmmm, I am working on a project called daiquiri that will re run Apache common
> logs against a cluster of web servers (or just one) and you can "multiply" the
> log times a factor (make every request twice or a hundred times)... It is
> available out of CVS at :pserver:anonymous@commedia.cnds.jhu.edu:/storage/cvs
> and the module is called daiquiri. If you don't know how to use CVS, I am
> afraid that you will have to wait for the first public release which isn't yet
> scheduled.
I am rather familiar with CVS, so I may try it. Thanks for the info.

Kind regards

Markus

Markus Neteler wrote:
> Funny thing is that:
> http://www.geog.uni-hannover.de/backhand/
> says: #ready/#used servers: 0/0
> Although the machine is running properly. Mhhh.
> Sorry to bother you here again, perhaps you have an idea for that.

It is the last question on the FAQ :-)

http://backhand.org/mod_backhand/FAQ.shtml#question13

--
Theo Schlossnagle
1024D/A8EBCF8F/13BD 8C08 6BE2 629A 527E 2DC2 72C2 AD05 A8EB CF8F
2047R/33131B65/71 F7 95 64 49 76 5D BA 3D 90 B9 9F BE 27 24 E7

On Sat, Nov 11, 2000 at 03:31:56PM -0500, Theo E. Schlossnagle wrote:
> Markus Neteler wrote:
> > Funny thing is that:
> > http://www.geog.uni-hannover.de/backhand/
> > says: #ready/#used servers: 0/0
> > Although the machine is running properly. Mhhh.
> > Sorry to bother you here again, perhaps you have an idea for that.
>
> It is the last question on the FAQ :-)
>
> http://backhand.org/mod_backhand/FAQ.shtml#question13
Oops. Sorry for that.

But if you are interested in background information as your are
working on that:

- in the first try of backhand, when it was working (before I
pained you with mails), the number of servers was correct (SUN)
- restarting apache does not help... Maybe I have to reboot
the machine.

Never mind. If this wrong numer does not hinder backhand to work,
it does not matter.

Thanks for your patience

Markus



>
> --
> Theo Schlossnagle
> 1024D/A8EBCF8F/13BD 8C08 6BE2 629A 527E 2DC2 72C2 AD05 A8EB CF8F
> 2047R/33131B65/71 F7 95 64 49 76 5D BA 3D 90 B9 9F BE 27 24 E7
>
> _______________________________________________
> backhand-users mailing list
> backhand-users@lists.backhand.org
> http://lists.backhand.org/mailman/listinfo/backhand-users

--
Dipl.-Geogr. Markus Neteler * University of Hannover
Institute of Physical Geography and Landscape Ecology
Schneiderberg 50 * D-30167 Hannover * Germany
Tel: ++49-(0)511-762-4494 Fax: -3984