[4] in 6.033-lab
Re: Recovering from server crashes
daemon@ATHENA.MIT.EDU (Constantine P. Sapuntzakis)
Thu Mar 6 23:23:33 1997
To: "Kevin 'Bob' Fu" <fubob@MIT.EDU>
Cc: 6.033-lab@MIT.EDU
In-Reply-To: Your message of "Thu, 06 Mar 1997 23:14:41 EST."
<9703070414.AA12782@gaston.MIT.EDU>
Date: Thu, 06 Mar 1997 23:24:02 EST
From: "Constantine P. Sapuntzakis" <csapuntz@MIT.EDU>
> In your email, you said to reference Tanenbaum on recovering from a
> server crash with at-most once semantics.
>
> But Tanenbaum says to return an error code. That is, no
> retransmissions. Is this right? What am I missing?
That's right. There's not much you can do, because the RPC protocol
doesn't know how much of the procedure call that the previous server
executed. So, it assumes the worst. That is, it assumes the procedure was
executed before the server crash but that reply was not succesfully
communicated. Unfortunately, the new server no longer has access to that
reply to resend to client. Nor can it rerun the procedure without
breaking at-most once semantics. So it's only choice is to return an
error.
-Costa
>
> --------
> Kevin E. Fu aka Bob the BobOp Athena OLC/RCC
> PGP key: finger fubob@snafu.mit.edu SIPB Member
