Re: Bug fixed ... Re: PROBLEM REPLICATED (finally!)! (Re: ipv4 connection problems)

[Bill Hacker <wbh@xxxxxxxxxxxxx>]

5J1QoR085625@xxxxxxxxxxxxxxxxxxxx>
In-Reply-To: <200504051901.j35J1QoR085625@xxxxxxxxxxxxxxxxxxxx>
Content-Type: text/plain; charset=us-ascii; format=flowed
Content-Transfer-Encoding: 7bit
Lines: 53
Message-ID: <4252f4e5$0$718$415eb37d@xxxxxxxxxxxxxxxxxxxxxxxxxxxxxx>
NNTP-Posting-Host: 218.253.82.16
X-Trace: 1112732901 crater_reader.dragonflybsd.org 718 218.253.82.16
Xref: crater_reader.dragonflybsd.org dragonfly.bugs:4043

Matthew Dillon wrote:

> :IIRC, that was the last major potential 'show stopper' to a 'good 
> :enough' for production box- potentially an even bigger problem with an 
> :MTA than with Apache.
> :
> :Is there anything else of that severity left? (that we know of?)
> :
> :Bill
> 
>     I'd like to nail down the X related panic report (which is probably not
>     related to X, just that X happened to be running when it occured),

I'm more inclined to think it is *very much* related to X (and/or GNOME 
2.10 ishness),
else more of us would have seen it.

> and
>     the TCP flooding report but they aren't show stoppers.    We are still
>     on course for a Friday release.  I'm going to start rolling test ISOs
>     tonight.

I've got a box 'targeted' for the first of those that reaches the public.

I will be out of HKG for Virginia and Europe 21 April - 8 Sept, so must
decide fairly soon whether I want to be admin'ing FreeBSD servers
or DragonFly servers from that sort of distance. (Some are spares)

DragonFly's 'just like old times' rapid ssh response, among other things,
has me leaning toward biting that bullet and converting all but a couple
servers to DragonFly before I go.

> 
>     The TCP flooding report could be related to the SMP TCP bug just recently 
>     fixed.  From Atte's report it could also be related to the packet filter
>     somehow.  There's not enough info on it yet to know either way.
> 
> 					-Matt
> 					Matthew Dillon 
> 					<dillon@xxxxxxxxxxxxx>

Have six BSD servers in that rack with public Internet frontside,
private LAN backside NIC's.

Can test at 100 Mbps on the internal, perhaps 'gang up' on the victim box,
as al are on the same 10/100 switch (frontside is split over two such).

Is the appropriate test suite the stress test you recently committed,
and will it compile under FreeBSD 4.XX --- or is something else
a better tool?

Bill