On Tue, Sep 18, 2012 at 12:39 PM, Bryan Beaudreault <[email protected]> wrote: > Looking closer at it I guess the flush and the IOException probably weren't > related. So the multi call to delete must have failed at the client (which > is good). It does seem very strange to me that the pattern always seemed > to be 1) start flushing, 2) fail with Cannot append. This happens a > handful of times. So we're still trying to find reasons for this loss of > data. >
Put up more log around the shutdown on pastebin of regionserver Bryan? St.Ack
