Thomas SMETS
Posts: 307
Nickname: tsmets
Registered: Apr, 2002
|
|
Re: Injecting fault in JVM
|
Posted: Aug 16, 2005 5:07 AM
|
|
Zero<sup>th</sup> :
I think your approach is definitively wrong... or at least should almost NEVER be used... You make clusters, use redundancy (hardware, software, power supply, networking, human, ...), etc... but at some point you have to stop.
If your system is capable of doing a hot node removal ... I think you are as far as most major telco expect. If you have to go further, you should then consider to have your system "voting"... but then you are also in considering buying very special hardware, maybe...
Anyway, if you consider this being part of system testing, I would strongly recommand you to firstly reconsider the following :
1°. See the possibility to measure your level of code coverage with your unit tests.
2°. Ensure that your in the big sftware layer, there are some places where you catch all Exceptions (Runtime ones, I mean).
Now as part of your system tests you may introduce some weird aspects that will generate emulations of those mistakes ... This would be AFAI can imagin a nice leap ahead... but you also have to be able to admit that your software has a price and then at some point, you have to stop testing what others have tested for you ...
\T,
|
|
Previous Topic
