Neil, I'm a little late to the game here, but if you provide details of the
problems you're experiencing, then I'll do my best to help out. 

ColdFusion configurations and environments *vary widely* among users. I
don't think that casting a broad net for specious reports, incomplete data,
and hyperbole will serve to solve whatever particular difficulty ails your
server, although there is certainly much to be said for a controlled
gathering of data "from the field" for careful analytic review. Some things
to consider might be the following:
I) Is this a server crash (jvm stops), a server hang (unresponsiveness with
live jvm), or a certain frequency of errors with neither a crash nor a hang?
The subject line says "Crash", but the message says "errors".
II) If its a crash and the jvm has exited and perhaps restarted, then:
        A) What is in the JRun log files *from the time of the crash*?
                1) java.lang.outOfMemoryError?
                2) is there a match for Sun bugs 5048441 are 5075468?
                3) how about GC logging or JRun metrics logging? What
                does that tell you?
        B) Does a hs_err_pidnnnnnn.log exist (where nnnnn is a processid) 
                anywhere under the JRun root or in the system's Temp 
                directory?
                1) what's in the hs_err_pidnnnnnn.log?  This is a HotSpot 
                crash log and it typically contains information about 
                what was being called at the time of the crash, and it
                typically contains a HotSpot Error Code at the bottom
                which can be used for Googling or searching Sun's Bug
                Parade.
                2) does the addition of -Xint to the jvm args list make
                a difference?
        C) What are the settings in the jvm.config file?  
                1) have there been any unusual settings added?
                2) have the settings been altered beyond the default?
                3) how suitable are the min and max heap sizes for
                your application(s)?
        D) Known issues (Technote 18325).  Are there any matches?
                a) be sure to use the 3.3 Macromedia drivers version.
III) If the server's jvm seems to stay up but the server stops 
handling requests, then:
        A) How are the principle server settings tuned?
                1) activeHandlerThreads (a.k.a. Simultaneous Requests)
                2) Timeout, the general one in the CF Admin
        B) Have you logged slow pages as per the CFAdmin logging page
        and weeded out obvious bottlenecks?
                1) if so, then begin to drill down into the pages reported
                to isolate what logical parts of the code to account for 
                the long run times.
                        a) use gettickcount() method of time tracking
                        b) use cflog, cftrace
                        c) use regular CF debug output
        C) Is a single server suitable to the amount of load and expected 
        throughput? IOW, if the server is tuned as best as possible, and
        the application code refined, then do you still need more?
                1) have you load tested to satisfaction?
                2) is there a need to scale up to JRun clusters across
                machines?
                3) is there a need for additional hardware load balancing?
        D) Have you performed thread dumps from the time of a hang to
        observe what code is executing at that time?  They are most
        helpful when taken 2 or 3 at a time about 15-30 seconds apart.
        E) Known issues (Technote 18325).  Are there any matches?
        F) Check out DevNet Performance articles such as Jim Schley's
        "Performance Under the Covers in ColdFusion MX 6.1"
IV) Unusually high frequency of errors without actually hanging or crashing
the server.  This is a fairly wide topic and specific to an installation, so
this would likely be tedious and specific to a given installation or
environment.
V) Have you contacted Macromedia ColdFusion support directly?  Those tickets
assessed to be verifiable bugs are closed at no charge (charges are reversed
or support packs incremented).  

I think Michael's list will cut me off here if it hasn't already, so
consider the problem objectively and try to provide as much relevant,
empirical evidence as possible.  Others will benefit from your analysis and
most importantly they will be more likely to help resolve it.

-Steven Erat
Macromedia ColdFusion Support

~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~|
Special thanks to the CF Community Suite Gold Sponsor - CFHosting.net
http://www.cfhosting.net

Message: http://www.houseoffusion.com/lists.cfm/link=i:4:185267
Archives: http://www.houseoffusion.com/cf_lists/threads.cfm/4
Subscription: http://www.houseoffusion.com/lists.cfm/link=s:4
Unsubscribe: http://www.houseoffusion.com/cf_lists/unsubscribe.cfm?user=89.70.4
Donations & Support: http://www.houseoffusion.com/tiny.cfm/54

Reply via email to