Thank for the encouraging feedback.
Your config looks OK, maybe you would also like to add a
connect_timeout, which works like prepost_timeout but when creating a
new backend connection.
Regards,
Rainer
[EMAIL PROTECTED] wrote:
Just as a courtesy in case anyone wonders the outcome of this thread:
I have tested the latest subversion development iPlanet connector
indicated by Rainer (compiled using Sun's compiler but gnu make - also
updated to link with Sun's compiler rather than ld .... as recommended in
the man page for the Sun compiler). This latest version with patches
including "un disabling" of connection pooling appears to work quite
happily in our test environment. I will roll it out to production when the
next stable release is done most likely.
Using the stable 1.2.23 release with just the "un disabling" "fixed",
compiled in the same way, in production has so far managed to hold it's
own without great issue. There are a number of what I believe are spurious
log entries regarding tomcat servers being unavailable, and only around 30
odd connections ever seem to get held open at any one time despite having
64 set in my workers.properties (included below - and hopefully not
misconfigured - the XX are intentionally to hide IP's). However, so far
this doesn't appear to have adversely effected things.
As an indicator, the site we run is fairly busy, with only a small subset
of content being served by the tomcat servers, but still we get around 7
million hits a day handled by our 3 tomcat servers.
worker.list=jknsapiworker
worker.host1.port=8009
worker.host1.host=XX.XX.XX.0
worker.host1.type=ajp13
worker.host1.lbfactor=1
worker.host1.socket_keepalive=True
worker.host1.socket_timeout=30
worker.host1.prepost_timeout=5000
worker.host1.reply_timeout=180000
worker.host1.connection_pool_size=64
worker.host1.connection_pool_timeout=600
worker.host2.port=8009
worker.host2.host=XX.XX.XX.1
worker.host2.reference=worker.host1
worker.host3.port=8009
worker.host3.host=XX.XX.XX.2
worker.host3.reference=worker.host1
worker.jknsapiworker.type=lb
worker.jknsapiworker.balance_workers=host1,host2,host3
Dale Roberts
Sony Computer Entertainment Europe
http://www.scee.com
[EMAIL PROTECTED]
09/07/2007 09:55
Please respond to
"Tomcat Users List" <users@tomcat.apache.org>
To
"Tomcat Users List" <users@tomcat.apache.org>
cc
Subject
Re: iPlanet / SunONE web server tomcat connector connection re-use
disabled
Hi Rainer,
Many thanks for the feedback.
To answer your question on stability, without the connection pooling being
"un disabled" the connector appears stable until placed under load at
which point it starts to flag all tomcat instances as "down", which they
are not. (I don't believe this is a result of over busy back-end tomcat
servers though based on some preliminary observations of number of
connections, tomcat logs, tomcat server performance etc). Unfortunately I
couldn't give you any figures on how much load is required (we have a
system under light load and a system under heavy load and nothing to "test
the middle ground"). However, with the connection pooling "un disabled" it
appears to work much better having run for most of the day on our heavy
load system without issue before being "rolled out" for operational
reasons not related to stability.
To answer your question about ongoing support for the iPlanet / SunONE
connector, from an ongoing support point of view I think that's really a
matter of how many people use iPlanet and how hard it is to maintain. That
said, I for one appreciate the fact there is a connector. The only other
option I can see is to configure one of the proxy forwarding modules of
iPlanet available for Sun, but I suspect that wouldn't facilitate
connection pooling and would thus provide lower performance and I had
trouble trying to download it anyway.
As to bugzilla and patches, I will do my best (I have signed up for a
bugzilla account today so I can open a bug to track this). However I feel
I should point out that I'm not a developer. My last proper C experience
was a university (and that was C++) over 10 years ago. So I know enough to
read some code and was able to find this "possible bug" but I fear I might
do more damage than good if let loose on source code "for real". I work at
the infrastructure level (more an end use of the connector). So apologies
in advance if I don't contribute to further updates and fixes but I will
try to provide the specifics of the update I made in suitable format in
the bugzilla report when I raise it.
Dale Roberts
Sony Computer Entertainment Europe
http://www.scee.com
Rainer Jung <[EMAIL PROTECTED]>
07/07/2007 13:33
Please respond to
"Tomcat Users List" <users@tomcat.apache.org>
To
Tomcat Users List <users@tomcat.apache.org>
cc
Subject
Re: iPlanet / SunONE web server tomcat connector connection re-use
disabled
Hi Dale,
thanks for digging into this. In fact the Netscape/Sun connector is not
in very good shape at the moment. The amount of work that will go into
it, to bring it into a better shape depends on people like you, that are
going to help us. The Netscape/Sun community seemed to almost
non-existant for many months and only recently a few people showed up
and asked questions about the connector.
Would you please open a bugzilla for your observation and add your
changes as a patch to it? This way it is more likely, that we will
remember this as an open point.
You might also like to check out jk from subversion and build the
Netscape connector from this version. I added some enhancements/bug
fixes which will be released with version 1.2.24.
What is your general impression about the Netscape connector? Is it
working stable for you? Do you think it is worthwhile supporting it in
the future?
Concerning your question: I think the missing connection reuse is just a
historic oversight. Enabling it needs some testing though in order to
ensure it runs well.
Regards,
Rainer
[EMAIL PROTECTED] wrote:
Hi,
BACKGROUND:
I have been using the tomcat connector for Netscape / SunONE web server
from the tomcat-connectors-1.2.23-src bundle available from the main
tomcat site. I have found that connections are not being re-used by the
connector. i.e. the connector opens a separate connection to tomcat for
each and every request it's configured to pass to tomcat and then closes
that connection after receiving it's response.
When I enable debug logging for the connector I see messages telling me
that connection re-use is disabled. I found reference to an option for
apache to disable re-use but nothing for Netscape. Looking at the source
code I find that all but the Netscape connector appear to have a
"s->disable_reuse = JK_FALSE;" statement in the "init_ws_service"
function
in the connector source code file (jk_nsapi_plugin.c for Netscape).
As a test I have added this to the Netscape connector and re-compiled
and
it now re-uses connections which is an obvious performance gain (I'm
assuming here as I haven't benchmarked anything). The implication is
that
someone either explicitly set re-use to be disabled for Netscape,
presumably due to problems not immediately obvious if re-use is
enabled,,
or someone changed the common code to make re-use optional but forgot to
enable it by default in the Netscape connector. I only suggest the
latter
as there is no immediately obvious indication that it is intentionally
disable in Netscape and the plug-in seems to work with it enabled after
some very brief testing.
QUESTION:
Can anyone confirm or deny whether the Netscape connector should not
have
connection re-use enabled as the other connectors do. Some overview of
any
reason would also be much appreciated.
Many Thanks in advance.
Dale Roberts
---------------------------------------------------------------------
To start a new topic, e-mail: users@tomcat.apache.org
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]
**********************************************************************
This email and any files transmitted with it are confidential and intended
solely for the use of the individual or entity to whom they are addressed.
If you have received this email in error please notify [EMAIL PROTECTED]
This footnote also confirms that this email message has been checked for
all known viruses.
Sony Computer Entertainment Europe Limited
Registered Office: 10 Great Marlborough Street, London W1F 7LP, United
Kingdom
Registered in England: 3277793
**********************************************************************
---------------------------------------------------------------------
To start a new topic, e-mail: users@tomcat.apache.org
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]
