Steve has a few entries about the problems he’s having with JRun connectors, and admittedly they’ve been quite a thorn for me for quite some time. I’ve pretty much narrowed the problem down to them being problems with connecting to JRun Clusters as connections to single instances seem to work fine (this is with IIS 5 and Windows Server 2003, by the way). This has been a long standing problem for me, so I’ve tried all kinds of installs from ColdFusion MX 7-7.02, with and without hotfixes, etc.
Here is the list of symptoms I’ve seen in general:
- IIS Hangs and does not return responses
- All kinds of entries in the connector log that look like this
- Restarting the IIS Admin and dependent services will temporarily clear-up the problem, and then it starts happening again
- It seems to happen more frequently when the machine has less load, but I have no emperical evidence of that to pass on
- If you connect to the instances directly via a browser, they are running just fine
The things I’ve tried without success / double checked:
- Tried the connector both with and without native OS memory allocation
- Made sure that no other services on any of the servers were using any of the same ports as JRun
- Made sure that all NIC’s and ports on the switch were configured for 100 Base T
- Installed IIS on an application server, and connected it to the cluster on the same server (to eliminate network problems as a possible cause)
- Installed the IIS Hotfix
The one work around to the problem is to delete the registered instances and cluster for the server, and just connect directly to one of the instances, so it seems to be some sort of problem with JRun Clusters. I have a hard time trying to figure-out what the problem could be given the large amount of time I’ve spent on troubleshooting this (even to the point of redoing server architecture to work around it), and the most frustrating thing is that clustering doesn’t even seem to work correctly when putting IIS on the same machine with the app server.
Has anyone else had problems with the JRun connectors when connecting to JRun clusters? Luckily I won’t have to deal with them for much longer given a new change in architecture soon which will allow them to be load balanced via a load balancer at the application tier as well instead of just being balanced at the web tier, but until then I’ll probably end-up removing all connections to JRun clusters and have IIS talk to a single instance instead since it seems to be a “known good” configuration.
By Claude July 14, 2006 - 2:48 pm
Did you try running on a non-Windoze server?
I manage a load balanced environment with 4 RedHat Linux Enterprise servers, each box has 2 JRun clusters, one of 2 and the other of 4 instances of CFMX7/Flex 1.5 — Yes that’s 24 hybrid instances.
Also check out Steven Erat and Brandon Purcell’s blogs.
By Brandon Harper July 14, 2006 - 3:11 pm
I haven’t tried the setup on Linux and the like, but at this point it’s not an option (and isn’t worthwhile given that this configuration will only be up for another 3 weeks or so).
And yep, I read their blogs and have an entry dedicated to Clustering CF and what not. I’ve been setting-up/tuning enterprise clusters for a bit short of 4 years now, have been doing sysadmin on Windoze, Linux, and FreeBSD for 9 years or so at this point, Linux certified, etc. and this is about the only problem which I haven’t been able to work around as far as CF/JRun configuration is concerned.
I’m curious to see if anyone besides Steve and I have had the exact same problem and/or knows a specific workaround for it since it’s very specific problem and I think I’ve exhausted everything I can think of to try.
By Steve Ray July 14, 2006 - 5:40 pm
Brandon -
Thanks for the post. I have now moved onto Apache 2.0.58. It is loading the connector properly, from what I can see in the apache error log. And, my individual apache virtual host error logs have nothing in them. I’ve followed some advice from Steven Erat’s blog on configuring mult virtual hosts, and yet, all I get in the browser window is the CF code displayed (not run). I’ll continue to work on this on Monday, so if you think of anything else, please pass it on. Thanks again.
By Alex Hubner July 14, 2006 - 7:17 pm
I’ve faced many problems with connectors both on Windows and Linux machines. Whenever you configure the connector to a Jrun cluster it became a instable and unpredictable sollution. Things get worst when you have a cluster with remote instances. Last time I’ve tried to use it was in a RHEL 4 with Apache 2.0.x with 6 Jrun/CF instances (4 on a machine and 2 on another, via a 1Gbps network). It does not work as it should, does not balance as it should and worst: hangs the entire cluster, cloging evertyhing. Now I’m only buiding a cluster with replicate sessions on and using a third load-balancer (which can be a dedicated appliance or a software) to forward HTTP requests to the instances itself (8301,8302,8303, etc…). The Jrun/CF built-in webserver seams to handle well a lot of traffic (you should tune it on the jrun.xml file) and the sollution is pretty fair and reliable.
Some free clustering softwares that I’ve cogitate as a bypasss to the Jrun/CF connector:
Sun WebServer – http://www.sun.com/download/pr.....d=446518d5
HAProxy – http://haproxy.1wt.eu/
Balance – http://www.inlab.de/balance.html
Pound Reverse Proxy – http://www.apsis.ch/pound
Pen – http://siag.nu/pen/
And of course, Apache mod_proxy (including mod_proxy_balancer).
By Brandon Harper July 14, 2006 - 7:28 pm
Steve,
No problem– I’m working on almost the exact same thing at the moment and have not had any luck with it for quite a long time and was reluctantly glad to see someone else with the same issues.
Alex,
Sounds very similar to my experiences. We’re in the process of moving to having a dedicated load balancer direct traffic in the exact manner that you mention. Thanks for the links too, I’ll check those out.
By My Other Beans » Blog Archive » How to Configure JRun Clustering with Apache July 16, 2006 - 1:43 pm
[...] Brandon Harper reported some long standing issues he has experienced with JRun connectors and IIS. Although this post doesn’t address this exact problem, it will provide guidance for those experiencing similar problems with Apache. [...]
By Karl October 6, 2006 - 11:00 am
One guy who I know has had serious problems over time with Jrun and CFMX is Matt Haughey of http://www.metafilter.com .. metafilter is cfmx, it’s hosted on a single instance server, and he’s had stability problems for years.
By How to Configure JRun Clustering with Apache | ~/my-other-beans March 4, 2008 - 4:14 pm
[...] Harper reported some long standing issues he has experienced with JRun connectors and IIS. Although this post doesn’t address this [...]
By Ryan Stille August 12, 2008 - 2:50 pm
What is this ‘connector log’ of what you speak? I’ve never heard of such a thing. I’m running into some issues with a site thats using the IIS connector to talk to a JRun cluster.
By Amir November 12, 2008 - 11:36 am
The connector log is the output from the connector plugin for apache. its a text file that details the information exchange between the http server (apache) and the application server (coldfusion). The log is the info from the connector (or man in the middle).
HTH
By How to Configure JRun Clustering with Apache — Claude Betancourt's Political Blog February 14, 2009 - 7:13 pm
[...] Harper reported some long standing issues he has experienced with JRun connectors and IIS. Although this post doesn’t address this [...]