[122] in Info-AFS_Redistribution
Re: urgent question: what happens when server goes down?
daemon@ATHENA.MIT.EDU (Liz_Hines@transarc.com)
Tue May 14 18:05:41 1991
Date: Tue, 14 May 1991 17:16:19 -0400 (EDT)
From: Liz_Hines@transarc.com
To: Info-AFS@transarc.com, ernest@pegasus.dsg.tandem.com
In-Reply-To: <9105131441.AA05564@pegasus.dsg.tandem.com>
Here is my understanding of what happens:
You have a binary in a volume that is replicated on both server A and
server B. You are accessing the the binary from the readonly volume on
server A. Server A goes down. Your client machine will continue to
use the pages that it has in the local cache. If it goes to swap in a
page that is not in the local cache, it will time out trying to talk to
server A (60 seconds), and then try get the info from server B. As long
as server B is up, it will manage to get the new page and proceed. The
process will not die.
A couple of notes here.
1) Our algorithm for tossing things out of the cache gives a higher
priority to chunks of executables in deciding what chunks stay in the
cache.
2) If you are accessing a binary in a readonly volume and you lose
contact with all of the readonly copies, it will not switch to the
readwrite volume.
Liz Hines
Product Support Manager, File Systems
Transarc Corporation
