[Mageia-sysadm] is the svn server down?

[Maarten Vanraes]

Op zaterdag 31 maart 2012 15:12:18 schreef Michael Scherer:
> Le samedi 31 mars 2012 à 12:51 +0200, nicolas vigier a écrit :
> > On Sat, 31 Mar 2012, Michael Scherer wrote:
> > > For the record, I see that 212.85.158.145 ( alamut default gateway,
> > > whose reverse is gateway-mageia.ielo.net ) is not answering to ping and
> > > on the other hand, the default gateway for zarb.org ( 212.85.158.1 )
> > > answer to ping.
> > > 
> > > And I am pretty sure that this gateway is after our switch from a
> > > network point of view, and so that this should answer fine if our
> > > switch is unplugged.
> > > 
> > > So I suppose that this device ( 158.45 ) do send network
> > > topology/routing with OSPF or RIP etc, and for some reason, it no
> > > longer announce the route for our network.
> > > 
> > > It could be caused by our switch being broken, but that would be rather
> > > strange, as I do not know any network equipment that would act like
> > > this ( doesn't mean that it doesn't exist, but usually, network is not
> > > that smart ).
> > 
> > According to gradator, this is normal if the switch is down. When the
> > port is down, OSPF automatically clear the route.
> 
> But what about the default gateway ? IIRC, that's not our switch ( as
> this is just a layer 2 switch ), so did it disappear on purpose once the
> network disappeared too ?
> 
> > > So what did IELO said exactly ?
> > 
> > The port where our switch is connected is down.
> > 
> > But even if the swich is not the problem, it would be a good idea to
> > replace it for a bigger one, to allow us to connect the arm boards on
> > it (instead of connecting them behind valstar). We also need to replace
> > one of the disk on the arm boards which seems to be not working.
> 
> Connecting them behind valstar was also to protect them until we
> properly secured them, since the password is still weak and easy to
> attack from bruteforce. And AFAIK, we didn't plan to give direct access
> to people, so that was useless and I think moving to a bastion model for
> ssh access would be a improvement in term of security for others
> builders too ( jonund, ecosse, and maybe fiona, depending how the backup
> are done ).
> 
> Also, just replacing the switch is not a improvement, since we will have
> the same issue if this switch break ( unlikely, but so was the current
> breakage, or the various hardware issues we faced each time we got there
> ) so what about trying to have a more redundant setup ?
> 
> Most, if not all, of our servers have 2 ethernet interface, so we could
> try to see if bonding could help ( in case of ethernet card failure ) or
> see if we can find a setup with 2 switchs ( one that doesn't cost too
> much, cause of course, cisco can solve the issue for us, we just cannot
> afford it ).
> 
> Can we afford one or two switchs that support it ? ( I lost the name of
> the current flavor du jour in term of bonding )

i think you're looking for distributed trunking? ie: having 2 cables in two 
separate switches, so if a switch fails, it stays working...

but those switches are quite expensive, iinm at least 2kEUR/switch

non-distributed trunking is alot cheaper, but that just means half the servers 
are down.

or we could just not bond and have 2 different networks over such servers. that 
also works for redundancy, but on l3 instead of l2 and alot cheaper...