Krzysztof Oledzki <olel@...> wrote:
>So, is it possible to fix bounding driver to drop RUNNING flag when
>all slaves are down? Anyway, is it OK that in such case kernel
>produces thousands of "bonding: Error: found a client with no channel
>in the client's hash table" messages?
As far as RUNNING goes, that should just be a matter of having
the bonding master device (the bondX interface) support netif_carrier
(which is inspected by dev_get_flags() to pass or not pass IFF_RUNNING
to the caller, which is typically SIOCGIFFLAGS). I'll see if I can whip
that up, the obvious way to do it is at the end of each link monitor
pass or add / remove of a slave check for live slaves, and set
As for the "no channel" message, that's not one I've seen
before. Its immediate cause is that somebody is sending an ARP request,
and the balance-alb code basically finds an empty bucket in its hash
table, which isn't supposed to happen. I assume from this that you're
running balance-alb mode, if not, then you really shouldn't see this.
I'll have to look into this a bit, I don't see an obvious error right
off. Assuming that I've identified the immediate cause, you'll get one
message per ARP sent.
-Jay Vosburgh, IBM Linux Technology Center, fubar@...