Skip to main content
Participating Frequently
February 6, 2026
Answered

Coldfusion sites experiencing significant delays

  • February 6, 2026
  • 3 replies
  • 59 views

We are running CF-2021 Update-22 with IIS on Windows 2019. Suddenly today all Coldfusion sites started experiencing significant delays. Response times have increased to between 5 and 40 seconds, while they used to be within 100-200 ms. There have been no updates or changes on the server as far as I can see. Restarting or upgrading connectors didn't help, CPU usage is low and there is plenty of RAM and storage. Http and css files are served immediately, and Coldfusion debugger reports show that pages are still loading pretty fast: Total Execution Time is less than 100 ms. I didn't find any signs of the DoS attack.

So the issue must be somewhere between IIS and Tomcat. In fact, isapi_redirect logs show more errors then before, like

[Thu Feb 05 21:16:16.555 2026] [10156:7684] [info] ajp_process_callback::jk_ajp_common.c (2288): current reuse count is 192 of max reuse connection 300 and total endpoint count 400
[Thu Feb 05 21:16:17.662 2026] [10156:13788] [error] start_response::jk_isapi_plugin.c (1293): HSE_REQ_SEND_RESPONSE_HEADER failed with error=87 (0x00000057)

Can anyone suggest how I should diagnose the problem?

 

    Correct answer Charlie Arehart

    So to be clear, on the page where debug output says it took only 100ms, the page rendering time in the browser was several seconds (like all those you are saying are a sudden common problem)? That’s important for us to know.
     

    And indeed, if it’s true, the next question is whether you find your browser’s “developer tools” feature (in its Network tab) confirms that there was indeed all that slowness in the serving of the CF page itself. You may well find it’s from something that the CF page pulls in as HTML, which the browser is then hanging up in serving.

     

    If indeed the problem is “on the CF page” (despite what your debug output is telling you), you would next benefit from any tool that monitors what’s going on within any or all CF requests. The most popular tool is FusionReactor (which has a free 14-day trial), and the next is the CF PMT (new since CF2018, but far more heavyweight if implemented on the same server as CF itself).

     

    I’ll note also (as some others will want to) that you are playing a game of Russian Roulette in still running CF2021, which got its last update in early Dec (it came out in late 2020 so is now EOL and will get NO more updates, including security updates). There was another CF update in January, which was NOT made available for CF2021, so you are now vulnerable to what it fixed (and your perf problems COULD be related to that, though not necessarily).

     

    Indeed, you may report by tomorrow that the problem is “gone”. Some problems do come and go, and while you don’t see a DDOS, it could just be excessive request traffic of a new nature from perhaps spiders or bots (or hackers or AI engines). Again, a monitoring tool can help spot that, in tracking all requests (or you could try to dig through your web server logs, which is tough given that they track ALL requests not just CF).

     

    Finally, I’ll say that if the problem remains and neither this note (or my subsequent replies) nor any thoughts from others helps resolve things, I help people solve such problems about daily via remote screenshare consulting. I can sometimes solve even such knotty problems in far less than an hour (my minimum time interval is just 15 mins). More on my rates, approach, satisfaction guarantee, online calendar and more are at carehart.org/consulting. But clearly I’m trying to help WITHOUT requiring that from you.

     

    These sort of problem hit a lot of people, and there is ALWAYS an explanation--often a very surprising one, and it’s generally easily solved. The challenge if finding the root cause.  That can be heard for us to guess at here, which is why direct access (looking at it together) is often far more effective. But we’re here to try, if this is the only way you may be allowed or prefer to seek help..

    3 replies

    BKBK
    Community Expert
    Community Expert
    February 6, 2026

    I’ve just had another idea. Though you mentioned there were no changes, Antivirus or EDR (End-Point Detection and Response) updates often occur automatically.

     

    If an Antivirus update started "inspecting" the AJP packets on the AJP port, it would cause exactly this kind of delay, without any CPU spikes. So check from the logs of any such security software (in any is installed) whether it was active at the time. If so, then temporarily disable the Antivirus or add an exclusion for the ColdFusion and IIS directories, as well as the AJP port.

    Participating Frequently
    February 6, 2026

    Reasonable idea, I will check the EDR and Antivirus logs and setting, at least what’s available.
    Thanks!

    Participating Frequently
    February 7, 2026

    There is nothing in the Antivirus logs. Strangely, the issue just disappeared, there are no delays for any of the sites anymore. Thank you for your comments and directions.

    BKBK
    Community Expert
    Community Expert
    February 6, 2026

    The errors in isapi_redirect logs are a good place to start the diagnosis. Apparently, connections are involved. So, could you please share the workers.properties file and the full text of the connector element <Connector protocol="AJP/1.3"> (from \runtime\conf\server.xml).

    Participating Frequently
    February 6, 2026

    Yes, here is the workers.properties from one of the 7 Coldfusion sites currently running:
    ====================
    worker.list=cfusion
    heartbeat_interval=30
    heartbeat_limit=90
    worker.cfusion.type=ajp13
    worker.cfusion.host=localhost
    worker.cfusion.port=8020
    worker.cfusion.heartbeat_servlet_path=/__cf_connector_heartbeat__
    worker.cfusion.connection_pool_size=400
    worker.cfusion.connection_pool_timeout=60
    worker.cfusion.max_reuse_connections=300
    worker.cfusion.monitoringsecret=660cebb9-7291-4185-acfa-39ad5831a7a1
    worker.cfusion.secret=fe3d3b10-1fd9-4bc6-963b-5a5563140757
    =========================
    <Connector  packetSize="65535" protocol="AJP/1.3" port="8020" redirectPort="8453" secret="fe3d3b10-1fd9-4bc6-963b-5a5563140757" maxThreads="1300" connectionTimeout="60000" tomcatAuthentication="false"/>
    =========================
    Thanks for reaching out!

    BKBK
    Community Expert
    Community Expert
    February 6, 2026

    @cfuser123321 , Thanks for your reply. There are some discrepancies in the workers.properties file and AJP connector element (assuming there is just one website using the workers.properties file):

    1. The values of worker.cfusion.connection_pool_size (400) and worker.cfusion.max_reuse_connections (300) are different. Optimally, you should set them to the same value. For example, using ColdFusion’s default
      worker.cfusion.connection_pool_size=500
      worker.cfusion.max_reuse_connections=500
    2. The maxThreads (1300) setting in the server.xml file and worker.cfusion.connection_pool_size (400) in workers.properties are vastly different. Optimally, they should have the same value. So, taking the first suggestion into account, you should use the setting
      maxThreads=500

    Remember to restart ColdFusion after making the changes.

    We can’t say for sure, but these discrepancies might be the cause of the issue. In any case, it is necessary for you to implement these changes for the long-term health and stability of your application.

    Useful reference: https://www.petefreitag.com/blog/tuning-tomcat-iis-connectors/

    Charlie Arehart
    Community Expert
    Charlie ArehartCommunity ExpertCorrect answer
    Community Expert
    February 6, 2026

    So to be clear, on the page where debug output says it took only 100ms, the page rendering time in the browser was several seconds (like all those you are saying are a sudden common problem)? That’s important for us to know.
     

    And indeed, if it’s true, the next question is whether you find your browser’s “developer tools” feature (in its Network tab) confirms that there was indeed all that slowness in the serving of the CF page itself. You may well find it’s from something that the CF page pulls in as HTML, which the browser is then hanging up in serving.

     

    If indeed the problem is “on the CF page” (despite what your debug output is telling you), you would next benefit from any tool that monitors what’s going on within any or all CF requests. The most popular tool is FusionReactor (which has a free 14-day trial), and the next is the CF PMT (new since CF2018, but far more heavyweight if implemented on the same server as CF itself).

     

    I’ll note also (as some others will want to) that you are playing a game of Russian Roulette in still running CF2021, which got its last update in early Dec (it came out in late 2020 so is now EOL and will get NO more updates, including security updates). There was another CF update in January, which was NOT made available for CF2021, so you are now vulnerable to what it fixed (and your perf problems COULD be related to that, though not necessarily).

     

    Indeed, you may report by tomorrow that the problem is “gone”. Some problems do come and go, and while you don’t see a DDOS, it could just be excessive request traffic of a new nature from perhaps spiders or bots (or hackers or AI engines). Again, a monitoring tool can help spot that, in tracking all requests (or you could try to dig through your web server logs, which is tough given that they track ALL requests not just CF).

     

    Finally, I’ll say that if the problem remains and neither this note (or my subsequent replies) nor any thoughts from others helps resolve things, I help people solve such problems about daily via remote screenshare consulting. I can sometimes solve even such knotty problems in far less than an hour (my minimum time interval is just 15 mins). More on my rates, approach, satisfaction guarantee, online calendar and more are at carehart.org/consulting. But clearly I’m trying to help WITHOUT requiring that from you.

     

    These sort of problem hit a lot of people, and there is ALWAYS an explanation--often a very surprising one, and it’s generally easily solved. The challenge if finding the root cause.  That can be heard for us to guess at here, which is why direct access (looking at it together) is often far more effective. But we’re here to try, if this is the only way you may be allowed or prefer to seek help..

    /Charlie (troubleshooter, carehart. org)
    Participating Frequently
    February 6, 2026

    Yes, you got it right, loading time is orders of magnitude longer than the Coldfusion execute time.. This is what surprises me most, and therefore, it looks more like communication trouble related to Tomcat.
    Thank you Charlie for a fast response!