This is the mail archive of the
gcc@gcc.gnu.org
mailing list for the GCC project.
Re: backup schedule on FSF machines
- To: Zack Weinberg <zack at codesourcery dot com>
- Subject: Re: backup schedule on FSF machines
- From: "Joseph S. Myers" <jsm28 at cam dot ac dot uk>
- Date: Sat, 3 Nov 2001 12:42:50 +0000 (GMT)
- cc: "Bradley M. Kuhn" <bkuhn at fsf dot org>, <gcc at gcc dot gnu dot org>, Gerald Pfeifer <pfeifer at dbai dot tuwien dot ac dot at>, <rms at gnu dot org>, <law at redhat dot com>, <overseers at gcc dot gnu dot org>
On Fri, 2 Nov 2001, Zack Weinberg wrote:
> There are several other things which we would like to know about
> before we consider shifting more operations onto machines maintained
> by the FSF.
All these questions also apply to the current gcc.gnu.org (overseers
CC:ed).
> Capacity considerations:
>
> - What is the maximum outgoing and incoming bandwidth from each
> machine potentially involved, to the network backbone?
>
> - How much disk space is immediately available on each machine? How
> many spindles is that split over? What is the practical peak
> throughput to each disk?
>
> - What kind of CPU load can each machine handle? How many concurrent
> CVS, web, and mail server processes?
- How many concurrent FTP processes? It would be good to be able to
have more than the current limits (for e.g. people trying to download
a new snapshot just after it is announced and before mirrors have
retrieved it). Are the current limits for machine load, or bandwidth
control? What ftpd implementation is currently being used?
- If it is found that more capacity is needed, how quickly would it be
paid for and added?
> Reliability:
>
> - Where are these machines physically located?
>
> - How old is the hardware being used?
>
> - How solid is the hardware being used?
>
> - What precautions have been taken to avoid system failures?
>
> - In the event of a system failure, how long would it take to restore
> service?
- In the event of localised problems, how long would it take to fix
corruption? (For the present system, past experience suggests several
months in many cases, relying on multiple volunteers rather than any
organised system.)
- What is the history of failures and recovery from them?
> Security:
>
> - How many people know the root password?
>
> - How many accounts exist on each machine?
- Of these, how many shell accounts?
> - How many of those are actively used?
>
> - What network services are available?
>
> - How long do security holes remain unpatched?
- How many people have physical access?
- How many people have physical access to the backups?
- What is the history of security failures and recovery from them?
--
Joseph S. Myers
jsm28@cam.ac.uk