[Piuparts-devel] Thoughts on piattiNG

[Dave Steele]

On Mon, Apr 15, 2013 at 7:29 PM, Holger Levsen <holger at layer-acht.org> wrote:
> Hi,
>
> On Dienstag, 16. April 2013, Holger Levsen wrote:
>> I'll send a reply to this (hopefully) with a URL for the "hw wishlist" to
>> Santa Claus^w^wDSA... :-)
>
> http://piratenpad.de/p/piatti-wishlist
>

Unless you know there is a lack of resources, I would start with a
negotiating position involving more slaves - particularly more
consideration for running multiple slaves on a slave VM.

I'd also provision resources, on a slave or elsewhere, for a backup of
the master configuration, logs, and counts.

I would have separated a public-facing reporting capability from the
master (which would also solve the backup issue), but that is water
under the bridge.

I would stay away from squid, as an initial position. It's not needed
if the mirror is local, and I've had problems with my Squid view
showing an inconsistent repository. The problem is not particularly
visible unless a 'missing' deb is highly dependent, but can be
transiently catastrophic. Also, if squid is local to the VM, now you
need to consider a greater level of inconsistency across processes.

It could make sense for you to own your own mirror. Good things can
happen if master knows he will be notified when the repository is
updated, and can manage the timing of the package deletion phase of
the mirror process.