| || |
Does SilkTest have a code-size limit (Running out of Processing Cycles)?
I have a MAJOR problem: When I run my code, the processing cycles go up to 100% at some point in the testing (beginning, middle, end - depends on the phase of the moon). I can run functions separately, and they work, but after running them a few times I need to reboot before continuing.
I am using a Win NT 4.0 system SP4, 350 MHz PII with 258 MB RAM. SilkTest is version 5.0 and I am using Netscape 4.51. The web application I am testing does allot with opening new browser windows for forms to be filled out, when logged in, etc. There is never more than 4 opened browsers at any one time. The problem usually presents itself as a page not loading (no errors, the page just doesn't load). Also note that there is never any problems with the application if Silk isn't being used.
The script I have, when it runs (VERY rarely), takes about 30 minutes. It consists of 35 files with a total of 15,972 lines of code. Here is a brief breakdown on the files:
AREA FILES LINES
Main Test Script: 1 790
Declaration Files 8 8711
Function Libraries 15 6201
Configuration Files 11 270
TOTALS 35 15,972
All the declaration and function files are called by a "Use" statement in 1 file. The main testscript then has a "Use" statement to include that one file. This ensures that all files are only included once. The configuration files are in an INI format and used to store frequently changing information.
I have tried the following with no success:
1. Running Agent remotely
2. Having nothing loaded by Silk and Netscape (after a fresh reboot)
3. Running each section as a testcase (works until I run all testcases via a TestPlan)
Each area of the code works, but when running all of them, or even just one area a few times, the processor maxes out and I have to re-boot. I have tried this on 3 different systems and I get the same problem on all of them.
333 Moffett Park Drive
Sunnyvale, CA 94089
Re: Does SilkTest have a code-size limit (Running out of Processing Cycles)?
PROBLEM IS SOLVED!!
(you won't beleive the cause)
Long... but worth the read...
The problem was that the CPU cycles were getting to 100%. At this point, the web page would no longer load or Netscape would go into an endless cycle trying to get the page. Silk would time-out after 10 minutes with the application not responding. I usually have to reboot after this since Netscape ate all the free and virtual memory and didn't properly release it.
Monitoring the system remotely using the NT Performace Monitor, I was able to see that SilkTest was only taking ~1% of the CPU cycles, Agent was taking ~3% - 7%, and Netscape was ~90%. However, if a script wasn't being ran, then Netscape was down around 10% - 15% of the CPU.
With this information, I assumed that since Silk queries the page about once every 0.06 seconds, it was the cause. I started playing around with Agent's settings, got the query to about once per second, which dropped the CPU cycles Netscape used down to about 70%. This allowed the script to run some, but it still failed. At some point, the CPU would still get to 100% and it bombed out.
ON TO THE REAL REASON IT FAILED.....
Did I mention I have three computers in my cube? All are connected to a 4-port hub. The network connection in the next cube wasn't good, so the system there was connected on the same hub as my systems.
Well, the person in the other cube left (he was an Intern). I unplugged the other system from the hub so I could use the port for somethign else. That same day, a developer asked if I could run my script agaist his system so he could verify if his changes broke the build. I explained that the scripts hadn't worked in months so there was no reason to even attempt it. He wanted me to try anyway since he made some global changes and didn't want to test the whole application.
I tried it - it worked! No problems at all. So, we thought it was just his system. Later in the day, tried it agaist the QE Lab systems - worked there too!
The network card of the PC in the NEXT cube was flakey. It was some how hosing Silk to the point where nothing would run!
For the past several months, I have been spinning my wheels trying to get Silk to work with our web application without much luck (was going good before that, then flacked out). About that time is when the Intern started and the PC went into the next cube. All this time, it wasn't Silk or our application or my code that was the problem - it was a bad network card - not even in the systems I was using, but in a DIFFRENT systems!
I couldn't beleive that that was the root of the problem, so I hooked the PC back up - well, the scripts died again, Netscape was back eating up the CPU cycles. Strange problem, but now I know what to watch out for.
333 Moffett Park Drive
Sunnyvale, CA 94089