On Wed, Jan 14, 2026 at 09:08:42AM +0000, Alisdair MacLeod wrote: > I did some more digging this morning and noticed that one of the interfaces > on veb0 had a weird number of failures considering there is nothing physically > attached to that interface. > > # netstat -I aq0 > Name Mtu Network Address Ipkts Ifail Opkts Ofail > Colls > aq0 1500 <Link> 00:e2:59:01:7e:35 0 0 93635 7414 > 0
can you tell if these Ofails are drops or errors by adding -e and -d to the netstat command? kstat aq0::: would also be interesting. > > So I removed that interface from veb0 and now the mbuf usage has stabilised. > > # netstat output > Jan 14 08:48:01 80072/80160 mbuf 2048 byte clusters in use (current/peak) > Jan 14 08:49:01 80126/80200 mbuf 2048 byte clusters in use (current/peak) > Jan 14 08:50:01 80176/80248 mbuf 2048 byte clusters in use (current/peak) > Jan 14 08:51:01 80286/80368 mbuf 2048 byte clusters in use (current/peak) > Jan 14 08:52:01 80393/80472 mbuf 2048 byte clusters in use (current/peak) > Jan 14 08:53:01 80441/80528 mbuf 2048 byte clusters in use (current/peak) > Jan 14 08:54:01 80504/80592 mbuf 2048 byte clusters in use (current/peak) > Jan 14 08:55:01 80556/80640 mbuf 2048 byte clusters in use (current/peak) > Jan 14 08:56:01 80609/80696 mbuf 2048 byte clusters in use (current/peak) > Jan 14 08:57:01 80620/80712 mbuf 2048 byte clusters in use (current/peak) > Jan 14 08:58:01 80620/80712 mbuf 2048 byte clusters in use (current/peak) > Jan 14 08:59:01 80621/80720 mbuf 2048 byte clusters in use (current/peak) > Jan 14 09:00:01 80622/80720 mbuf 2048 byte clusters in use (current/peak) > Jan 14 09:01:01 80619/80720 mbuf 2048 byte clusters in use (current/peak) > Jan 14 09:02:01 80620/80720 mbuf 2048 byte clusters in use (current/peak) > Jan 14 09:03:01 80620/80720 mbuf 2048 byte clusters in use (current/peak) > > I do remember now that this interface aq0 had some issues previously > which is why nothing is attached to it, but I did not consider that it > could still be consuming resources. I???m not sure if it is a hardware > issue but I will try to trigger the previous interface and then start a > new thread to investigate that as it looks like the PPPoE issues were > just a symptom of this misbehaving hardware. yes. like sthen said, i think you're leaking mbufs, which are the kernels representation of packets. this eventually makes it hard for the system to communicate on the network by making it impossible to do things like get packets to send pppoe keepalives with. > Thanks for all the help with the pointers of where to look for culprits. agreed, thanks stuart.
