mbuf_cluster (FAIL SLEEP)

Sat May 24 07:40:59 UTC 2014

> Yes, this is normal and it doesn't mean what you might expect.  It's a
> generic failure counter, not an allocation failure counter.  eg: if an
> object that was just freed fails to fit in a per-cpu free items cache it
> counts as a "FAIL".

Hello, thank you for answer Peter.
But what do think about this:

# netstat -an |wc -l
# netstat -an | grep -c TIME_WAIT

kernel stops showing realtime LA: show normal top, but no changes at first line
load averages:  2.36,  2.32,  2.28
                 ^	^	^
                 |	|	|

sysctl net.inet.tcp.msl and tcpdrop do not help,
after killing TIME_WAIT this connection state start grow again till 27800.

I have to reboot servers couse some apache process goes to <defunct> state
and stop responding.
Normal traffic and no ddos at this moment.

Again, this problem (netstat & LA) happen at the same time.....

More information about the freebsd-net mailing list