Re: Dom0 xvif mbuf issues

[Harry Waddell <waddell%caravaninfotech.com@localhost>]

On Fri, 5 Oct 2018 22:58:50 -0400
Thor Lancelot Simon <tls%panix.com@localhost> wrote:

> On Thu, Oct 04, 2018 at 05:01:21PM +0200, Manuel Bouyer wrote:
> > 
> > Maybe you should try reverting
> > kern.sbmax=1048576                   
> > net.inet.tcp.recvbuf_max=1048576     
> > net.inet.tcp.sendbuf_max=1048576     
> > 
> > to their default values.  
> 
> At least, the sbmax value needs to be slightly larger than the
> socket buffer sizes, because there's some overhead.  If you
> manually set sendspace or recvspace to exactly the sbmax value you
> will see problems quickly.  I don't know what happens with the
> autotuning code, though.
> 

Increasing sbmax seems to have helped, but I did suspend a few vms as well, 
so it's not conclusive yet. 

What I noticed was that before the previous hangs, the number of calls to the 
protocol drain routines would spike and then start growing over the course of about 
40 min until it would crash. 

Now, with 

kern.sbmax=4194304
net.inet.tcp.recvbuf_max=1048576
net.inet.tcp.sendbuf_max=1048576

it's been up 11 days and 

xen-09# netstat -m
514 mbufs in use:
        513 mbufs allocated to data
        1 mbufs allocated to packet headers
140 calls to protocol drain routines

it's still humming along. So far, so good. Thanks for the suggestion.


HW