Coda File System

Re: Replicating servers/SCM/cache size

From: Jan Harkes <jaharkes_at_cs.cmu.edu>
Date: Mon, 31 Jul 2000 14:21:10 -0400
On Thu, Jul 20, 2000 at 04:58:35PM -0400, Michael Sherman wrote:
> 1. Can SCM be moved from one server to another?

At the moment the other servers are only missing /vice/vol/VRList, if
you copy this file to another server it can completely take over all SCM
functionality.

>    Is SCM a single point of failure in a Coda system?
>    Can the rest of the system continue to operate when SCM fails permanently?
>    If so is there any data loss?

Everything except for creating/purging volumes and adding/removing users
will work fine even without the SCM. The only data loss would be that
VRList file, we intend to move it from /vice/vol to /vice/db, so that it
will become replicated across all servers as well.

> 2. I have come across a few suggestions as to the size of the cache on
> the client. They range from 20M to hundreds of megabytes. In my
> particular case client disk size is limited. Will they be able to
> operate with small caches (1 or 2 Meg)? Is it a requirement that a
> single file must fit into the cache? Are there any consequences?

All opened files which are held open by applications at the same time
must fit completely in the client cache.

Jan
Received on 2000-07-31 14:35:54