Ask Ben: Running ColdFusion Asynchronously - Caveman Style
Posted November 30, 2007 at 2:45 PM by Ben Nadel
Looking to run a CFM template asynchronously - aka: run it in the background. Gotchas:
* Must work under MX 6.1
* I can't use an event gateway
* Can't use CFSCHEDULE (unless you know of a way to post data to the URL)
This reminds me of that old Polly-O String Cheese commercial where the guy walks into the pizza parlor and says something like, "Give me a pizza, hold the tomato sauce, hold the crust". What we are trying to do here is run an asynchronous task without all the fixins that make ColdFusion good at this.
Aside from the fact that I am not sure what bullet point #3 means, the only way I can think of doing this is to create a ColdFusion page that calls itself recursively using the CFHttp tag. This has all kinds of pitfalls which make my stomach feel uneasy, but in this sample code I am trying to put in some effect minimizers that stop this idea from totally crashing your machine.
- <!--- Set file path. --->
- <cfset strPath = ExpandPath( "./log.txt" ) />
- <!--- Set safety file path (used at bottom). --->
- <cfset strSafetyPath = ExpandPath( "./run.txt" ) />
- Append the file hit. This is where you would normally
- do your processing, but for our testing purposes, we
- are just going to log the hit to a file.
- output="#TimeFormat( Now(), 'HH:mm:ss.L' )#"
- Sleep the thread for a little bit to give the
- server a break before this starts to demand
- resources once again.
- <cfset objThread = CreateObject(
- <!--- Sleep for 2 seconds. --->
- <cfset objThread.Sleep(
- JavaCast( "long", (2 * 1000) )
- ) />
- Before we call the page again, let's put in a
- check to make sure that we don't run FOREVER. We
- will only re-call the script if this file exists.
- <cfif FileExists( strSafetyPath )>
- Since this file exists, re-run the page using
- CFHttp with no wait time (or minimal wait time).
- If you ever want to kill this file, just delete
- that file.
- Your page has been re-launched in the background.
- If you want to kill this process, simply delete
- this file:
- This file cannot re-run until this file exists:
- <!--- Ouptut safety file. --->
- <cfset WriteOutput( strSafetyPath ) />
First, the help minimize some of the load on the ColdFusion server, I am asking the thread to sleep for 2 seconds before it re-launches itself using CFHttp. The second safety method is that this page will ONLY re-launch itself if the "run.txt" file exists. This file does nothing but server as a boolean flag; this is there so that you can delete or rename the file (run.txt) if you want the original script to stop running. This way, you could reasonably write another page that would conditionally rename the run.txt file based on URL params such that you could create a way to turn this script off using another script.
The caveat here is that you have to launch the first task manually. Once it gets going, though, it takes care of running itself.
Running this script, our log.txt text file ends up looking like this:
This whole thing makes me feel really uneasy. I am putting this out there because it was asked of me, but I really don't like it :) What might be a better methodology is to just build a simple HTML page that makes an AJAX call to the page every X number of minutes. The caveat there is that you need to keep that browser running, but at least it feels much safer to me.
regarding bullet point #3:
You cant post variables directly from a scheduled task. But your scheduled task could call a cfm page that itself does a cfhttp call. In that cfhttp call, you post as many vars as you want.
You're talking about a "POST" method right? In the user's question I was thrown by the statement "post data to the url". To me, this just sounds like query strings, which should be doable via a scheduled task (I am pretty sure). So maybe that is what they meant - posting to a URL as in (POST vs. GET).
Don't forget about using cfheader and status code 204! If thats the first thing that gets processed in the target 'async' template, the calling page will move right along after receiving a response and the target template will continue to process in the background.
I don't have tons of experience with it, but have verified it works in CF 6.1.
That is very interesting. I will do some experimenting with the 20X status codes. Thanks for pointing that out.
Mark's AsyncHTTP is very cool; it is one of the first things I found a while back when looking for that kind of functionality. Cool that it works CFMX6.1. Of course, this is basically doing the same thing as the CFHttp (I think) in terms of thread creation. The main difference is that there is no "Timeout" hack being used or required.
Depending on what it's being used for you could also use a hearbeat type plugin with jQuery.
Alternately, you could refresh the page to itself with a meta refresh. Low-tech and utterly simple.
Before CFThread came along I used a cfm page to do cfexecutes that would use wget at the command line to call the cfm pages to run asyncly.
All I can say is, thankfully CFThread came along and made our lives easier (and safer).
One big problem with this that I can see is that it uses up one of those concurrent requests in the pool. So if you have the server set to only allow 6 concurrent requests, after doing this, you're only allowing 4 (on average since you'd spawn and wait with two requests while this thing was running).
Most definitely. This is certainly not something I would recommend. But, it was the only server-side way I could think of making this work with such limitations.
This solution is awesome!
We were using the solution with manually creating schedule task which would run in 5 seconds and perform bunch of code once and then deleted itself.
We had to sync some data from out app to some other app that resided on different server and had some shitty db. Update call to their app was taking from 10 - 20 seconds. WAY TOO MUCH for us.
So we called the sync code in the background and continue to work with our app like everything was OK. We build log mechanism to control any failed syncs.
Anyway, cfhttp is way cleaner solution than cfschedule! BRAVO!
The simplest solution is probably the best one.