| Re: [epp-dev] Adding a error log collector to Mars milestone builds? |
The first thing that comes to mind is network traffic. Working with it for two days, it produced ~300 HTTP Post requests with several kilobytes of data each. If we assume that there will be a couple of hundred developers using that feature in the first milestone builds to report errors, this may sum up to: 200 developers x 150 requests / day = 30.000 requests per day. 30.000 requests per day x 10KB data = "" traffic. Having said this, since many errors look quite similar, the error log reporter can be optimized to send only „new" stacktraces in the future and thus reduce the network traffic. Second thing that comes to mind is disk space. For the time being, a few GB of disk space should suffice. At the moment the data is sent to recommenders.eclipse.org, a virtual machine Denis provisioned some time ago. It’s not a very powerful vm but from hardware perspective I’d say, I’m good to go for the first milestones. But I wonder whether making the error logs accessible as they are today is okay for EMO. See [1] for a collection of stacktraces I sent in the past 2 days. AFAICT, there is at least two log statements in JDT that log the source code of a whole file in the case of an error. Other statements contain names of files and ids (which cannot be filtered easily). Marcel Am 12.08.2014 um 13:52 schrieb Mike Milinkovich <mike.milinkovich@xxxxxxxxxxx>:
|