Bill

The uptime on both systems is 3 days or so.  The error counters on both 
machines for the pfsync interface is zero.

pf sync is a VLAN - it shares the same physical line as the LAN Interface 
which is also showing up with zero errors, as is the physical interface they 
are carved from.  I am using em(4) interfaces, Intel chipset.

I have switched to aggressive on the slave - not surprisingly the state usage 
has nose-dived.  However, I am not sure what impact that would have on 
traffic to the site.

I am a bit puzzled at how many states are in time-wait.  I must say that I 
thought the theory of stateful filtering was that the state would be deleted 
once the firewall had seen the fin2 go past.  I guess that aggressive just 
cleans up faster.

/Peter

On Saturday 27 May 2006 13:43, Bill Marquette wrote:
> Just a wild ass guess at this point I'm afraid.  Any chance you've got
> some packet loss on the sync interface, or really really crappy nics?
> It kinda sounds like the state deletion notices aren't always making
> it across.  The nice thing about state updates and why they're still
> obviously working is that if the secondary machine sees a state update
> for a state it doesn't have, it requests full info on that state and
> adds it.  So over the course of a tcp conversation it's going to have
> multiple opportunities to add state - not so for deletions.  In the
> meantime on the secondary, you could try setting the state
> optimization to aggressive - it might help a little, but I'd check
> your error counters on the sync interfaces and see if they're
> climbing.
>
> --Bill
>
> On 5/27/06, Peter Curran <[EMAIL PROTECTED]> wrote:
> > Bill
> >
> > I identified this problem a few days ago when asking about the sizing of
> > state table entries.  I have now had time to study the issue over a
> > longer period of time.
> >
> > The site I am working with is pretty busy - they typically have around
> > 10,000 punters on-line during the week:  The site provides price and news
> > information for commodity markets.  This translates to around 30-40K
> > states on the master firewall.  The slave is invariably showing
> > significantly more states in use than the master - typically around 70K. 
> > The discrepancy creeps up over time so that after a week or so, when the
> > master is showing 35K the slave is showing 95K (the max is set to 100K).
> >
> > If I reset the state table on the slave, it just starts off roughly in
> > sync with the master and then builds up gradually.
> >
> > I am not sure what is going on here - have you seen or heard of a similar
> > problem before?  Is there anything I can do to analyse the system?
> >
> > /Peter
> >
> > --
> > This message has been scanned for viruses and
> > dangerous content by MailScanner, and is
> > believed to be clean.
> >
> >
> > ---------------------------------------------------------------------
> > To unsubscribe, e-mail: [EMAIL PROTECTED]
> > For additional commands, e-mail: [EMAIL PROTECTED]
>
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: [EMAIL PROTECTED]
> For additional commands, e-mail: [EMAIL PROTECTED]

-- 
This message has been scanned for viruses and
dangerous content by MailScanner, and is
believed to be clean.


---------------------------------------------------------------------
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]

Reply via email to