I've had a little bit of weird behavior. I am opening a scanner in the configure method of a Map task to load a simple little in-memory map (I'd love to this with in-memory column stores, but that's another story ;-).
Anyway, the scanner iterates over 4 million rows or so and I've verified next is being called multiple times per second. However, at the same record every time (after around 6 minutes of iteration), the scanner throws an UnknownScannerException in the log (I see it in the log), but rather than throw the exception to the Map task, next just returns false and the loop ends (making it exceptionally difficult to detect if the scanner finished scanning or not). Increasing the regionserver timeout has most definitely solved the issue, but shouldn't calling next reset the timeout anyway? Such a frustrating problem!
