bridgescript needs a bigger timeout than 3s on some bonded setups

[Andrew McDermott]

The bridge script has this comment:

 # On configurations that have bonds in 802.3ad mode there is a
 # race condition betweeen an immediate ifdown then ifup.
 #
 # On the h/w I have a 'sleep 0.1' is sufficient but to accommodate
 # other setups we arbitrarily choose something larger. We don't
 # want to massively slow bootstrap down but, equally, 0.1 may be
 # too small for other configurations.

 for s in stanzas:
     if s.is_logical_interface and s.iface.is_bonded:
         print("working around https://bugs.launchpad.net/ubuntu/+source/ifenslave/+bug/1269921")
         print("working around https://bugs.launchpad.net/juju-core/+bug/1594855")
         print_shell_cmd("sleep 3")
         break

One some systems this is now not enough; it has been reported by <email address hidden> that they need at least 10s on their systems.

We can obviously raise the current limit by 1 order of magnitude, or better yet make it a parameter that can be passed in via bootstrap/deploy.

[Andrew McDermott (frobware)]

WIP tree - https://github.com/frobware/juju/tree/2.1-dynamic-bridges-fix-lp1657579

[John A Meinel (jameinel)]

This has landed in 2.1 as net-bond-reconfigure-delay
Defaulting to 17s