We have observed Journal Corruption while replaying the journal log.
I could be missing something but I don't see we have clear indication of
where
journal file ends.
In the Journal.java:scanJournal() we have multiple breaks but mostly
we assume either the journal file ends in the right place or t
JV,
As the machine/disk/process may crash/halt unexpectedly during writes we
can't do many assumptions, the only way is to accept an entry in case of a
valid read and consider that we reached the end of the journal in case of a
partial read or garbage.
In your case it is possible that the reader r
Correct. We can't just ignore it. We needs to be a proper EndOfEntryLog.
How did it work until now? I wrote a simple test to reproduce this. This is
super surprising to me, who we haven't hit it so far?
Filling the file with negative numbers reproduces this problem very easily
by running BookieJou