Hi Todd,
I think Arpit's test method is incorrect. we cannot block port 8020 to
simulate active NN down. because ZK session is live and NN process is
running at the same time.
so when unblock 8020, NN1 think himself still is active.
On Sat, Feb 8, 2014 at 3:47 AM, Todd Lipcon wrote:
> Hi Ar
Hi Arpit,
The issue here is that our transaction log is not a proper "write-ahead
log". In fact, it is a "write-behind" log of sorts -- our general
operations look something like:
- lock namespace
- make a change to namespace
- write to log
- unlock namespace
- sync log
In the case of an active
Hi Arpit,
In Your case, you blocked requests only on 8020 port. But ssh was reachable
right?
Have you configured fencing method? Such as "sshfence"
If you have configured, then previous ActiveNN should be killed before making
next one Active, Else shared storage needs to handle single writer m
