Re: mbuf cluster leak?

[Edgar Fuß <ef%math.uni-bonn.de@localhost>]

> So I'm making some progress on this.
But it still keeps hitting me.

After some quiescent days, the day before yesterday, mclpl suddenly started to 
ramp up, first at 12.5k/h, then 16.8k/h and finally at 23.3k/h.
At that point, half of my NMBCLUSTERS were used up, and I destroyed the vlan.

> I'm now running a kernel with MBUFTRACE (thanks MvE for the hint).
> What I see with netstat -mss is an abnormal increase in one of my vlan's 
> tx allocations.
That again showed an insane amount of tx allocations on that vlan (and 
"mbufs allocated to data").

> When I destroy that vlan if (and re-create/-configure it), that drops down 
> to approximately zero; but running a program that grabs all available 
> memory then doesn't lead to a substantial amount of Reclaim-s for mclpl.
I did that again, with two results:
1. The tx allocation went away, but "mbufs allocated to data" didn't.
2. The mclpl allocation stopped to rise and remained constant.

> I'm still wondering whether there's a way of inspecting the contents 
> of those mbuf clusters. I strongly suspect that looking at the contents 
> will give a strong hint at what's going on.
That question still holds.