Hi Robert,
Duly noted. I will pass your comments on to our tech folks. Thanks!
Brian Humes
Director, Interactive
JohnsonRauhoff
269-428-9257 (direct)
269-428-3377 (main)
269-428-3312 (fax)
www.johnson-rauhoff.com
[email protected]
On Oct 20, 2009, at 2:01 PM, Robert Garcia wrote:
There is also the consideration of whether you are only load
balancing witango servers or web servers as well. You can have one
webserver (apache/IIS) running with any number of witango servers in
a load group, and you can also have multiple webservers and multiple
witangos. You can have as many webservers as you do witango servers,
but not more.
The configuration of 1 webserver to many witango servers is simpler
to configure but does leave you with a single point of failure. If
you one webserver fails, your witango group will NOT receive
requests. If a witango server fails, the group should then not send
requests to the bad witango server, as you would expect, and as Mr.
Schubert has confirmed. However, you should be aware that it doesn't
always work. If you have a hardware failure on a witango machine, it
will work, the others will work fine. If one witango machine crashes
hard, you should also be fine, but there is a state that witango
servers can get into, where they crash in a way that just HANGS, and
doesn't recover, and in this state, they can act to the group like
they are ready for requests, but instead just hang on every request,
and the witango process on this server must be killed. In every
witango installation I have worked on, this will happen, and you
should prepare for it. I have worked on systems where this happened
all of the time, and then after working with them did it very
rarely. This is done by looking at the tafs and code and looking for
certain methods and algorithms that after much experience you find
can have a greater tendency to put witango into this state. An
example is a heavy use of certain types of external actions. These
can get witango in this state, which you want to avoid.
Another issue to consider is that even though witango should fail
over, and also in those moments where you must kill a witango
process if it hangs, any sessions on that witango system will also
be lost. This means that the users on that system will lose their
shopping carts or whatever other user variable sessions you have
going. Something to keep in mind.
If you go to a many webserver - many witango configuration to
eliminate the single point of failure for the web services, you will
have to employ some form of DNS load balancing. The simplest being
round robin dns. Round robin is great, but requires human
intervention to remove a dns entry if a webserver goes down, and a
few mins to propogate depending on the record's time to live. This
is usually ok, because web servers RARELY go down. There are other
options like hardware or software load balancing that can be setup
to failover without intervention.
--
Robert Garcia
President - BigHead Technology
VP Application Development - eventpix.com
13653 West Park Dr
Magalia, Ca 95954
ph: 530.645.4040 x222 fax: 530.645.4040
[email protected] - [email protected]
http://bighead.net/ - http://eventpix.com/
On Oct 20, 2009, at 7:41 AM, Brian Humes wrote:
Hi all,
I have a web server (Windows Server 2003 R2, IIS 6, Witango 5.5
Standard, MS SQL Server 2005) that is serving a lot of traffic. We
think its time to add a second server. Here's what we'd like to
accomplish:
1) When both servers are running, balance the load so each server
is getting roughly half the traffic.
2) If one server goes down for some reason, automatically shift all
traffic to the good server.
3) When a user hits one server, they stay on that server until the
session is completed (I think Witango does this automatically?).
I've gone through the server installation documentation and found
nothing about making this work. I did do some searching through the
archives and found that Witango is (it seems) capable of this, I
just don't know how to do it.
Could someone please point me in the right direction? Any help
would be greatly, greatly appreciated. Thanks!
Brian Humes
Director, Interactive
JohnsonRauhoff
269-428-9257 (direct)
269-428-3377 (main)
269-428-3312 (fax)
www.johnson-rauhoff.com
[email protected]
________________________________________________________________________
TO UNSUBSCRIBE: Go to http://www.witango.com/developer/maillist.taf
________________________________________________________________________
TO UNSUBSCRIBE: Go to http://www.witango.com/developer/maillist.taf
________________________________________________________________________
TO UNSUBSCRIBE: Go to http://www.witango.com/developer/maillist.taf
