Thank you for the responses. I am still not sure on the answer to my
second question.

>> 2. How is fault tolerance handled for an RS with a coprocessor loaded? Will
>> other servers load that coprocessor if the original RS crashes? If yes,
>> will the HLog be replayed with the coprocessor already loaded?

I will try to give an answer myself and maybe someone can correct me
if I'm wrong. I am referring here to the case when coprocessors are
loaded only for a certain table (loaded from the shell for example).
So essentially fault tolerance is handled only by distributing the
regions + HLog from the crashed RS to the other RSs. If that is the
case then new coprocessors will be instantiated when the new regions
are opened on the other RSs. However if the original RS crashed while
running an RPC call on an endpoint coprocessor, I don't see how this
can be handled without manual intervention from the client (by
restarting the whole job maybe). Can anyone confirm this?

Thanks,
Sever

Reply via email to