|
Yo guys,
> > Mhh, I don't know, what do others like e.g. Horms, Wensong and Julian think?
> > It's affecting 4 lines in the kernel and 14 lines in ipvsadm :)
>
> Some ideas for you:
>
> -Z -t/u VIP:VPORT - clear VS stats
> -z -t/u VIP:VPORT -r RIP - clear RS stats
> -Z/z - clear all stats
> -C/-D/-d ... - clears all/some stats
Cool, haven't thought of it that way, but I like it.
> What is the useful set of features?
All except -C/-D/-d IMHO.
> In the kernel-user interface this can be achieved using wildcards
> for the vs and rs fields in the user structure.
>
> I assume -Z is for humans. Kernel 2.4?
all kernels :)
> > We should maybe write something about cycling counters. This of course is
> > not handled. I mean when exceeding the 64bits we start over by 1 IIRC and
> > then we should set a flag to 1 to know that we already had 2**64 packets.
>
> Who needs such flag? It takes years to set :) Or may be after
> some years it will take months :)
Hey Julian, I've deployed load balancer with an uptime of 495 days (still with
the negative counter bug, persistency had to be disabled :). Of course I was
more kidding when I mentioned that. It definitely takes some time to exceed the
2**64 packets :)
Best regards,
Roberto Nibali, ratz
--
mailto: `echo NrOatSz@xxxxxxxxx | sed 's/[NOSPAM]//g'`
|
