requests for mbufs denied

[Kazuaki Oda]

Gabor MICSKO wrote:
> Hi!
> 
> I have a relative high traffic server, running Apache, MySQL and Drupal.
> With FreeBSD 6.0 and 6.1-PRERELEASE i got some distressing "netstat -m"
> outputs.
> 
> Can anybody explain for me what does this message mean exactly?
> 
> "16064849/9164254/9384500 requests for mbufs denied (mbufs/clusters/mbuf
> +clusters)"
> 
> And what can i do with this?
> 
> Full "netstat -m" output:
> 
> $ netstat -m
> 445/695/1140 mbufs in use (current/cache/total)
> 407/255/662/65536 mbuf clusters in use (current/cache/total/max)
> 407/237 mbuf+clusters out of packet secondary zone in use
> (current/cache)
> 0/0/0/0 4k (page size) jumbo clusters in use (current/cache/total/max)
> 0/0/0/0 9k jumbo clusters in use (current/cache/total/max)
> 0/0/0/0 16k jumbo clusters in use (current/cache/total/max)
> 925K/683K/1609K bytes allocated to network (current/cache/total)
> 16064849/9164254/9384500 requests for mbufs denied (mbufs/clusters/mbuf
> +clusters)
> 0/0/0 requests for jumbo clusters denied (4k/9k/16k)
> 33/964/6656 sfbufs in use (current/peak/max)
> 0 requests for sfbufs denied
> 0 requests for sfbufs delayed
> 56067 requests for I/O initiated by sendfile
> 5500 calls to protocol drain routines
> 
> 
> Sorry for my bad english.
> 
> Thank you!
> 
> 

In my case, that happens when the number of free pages become less 
than vm.v_free_min.  In such case, we drain cached buckets from zones 
(including mbuf zone) to the system to get more free pages, and 
disable bucket allocation.  As the result, we cannot get a bucket when 
freeing a mbuf to the zone and just do internal free, counting up the 
number of allocation failure, i.e. the number of "requests for mbufs 
denied".

I don't know whether this (as to counting up the number of failure in 
such case) is expected or not.

--
Kazuaki Oda