Re: NetBSD 5 partly freezing, may be related to 802.1Q

[Fredrik Pettai <pettai%nordu.net@localhost>]

Hi iMil,

On Jun 9, 2012, at 12:48 PM, iMil wrote:
> Hi Fredrik,
> 
>> Actually, we see a similar problem with a Dell R610 server, without using 
>> 802.1Q
>> But we've configured and use 3 of 4 interface's on the internal broadcom 
>> quad-card.
>> 
>> I also upgraded the server to NetBSD 6.0_BETA2, but the problem still comes 
>> back...
> 
> That's what I was about to do...
> 
> Just to try to reduce the scope, what is your server used for? what
> features do you use? Mine is a gateway, I have:

[...]
We had some features enable, but I remove those and used a GENERIC kernel to 
see if it still fails, and unfortunately it still does. 
I think the last kernel working fine was 5.1_GENERIC, and just 2/4 network 
interfaces used (but I'm not hundred procent sure).
btw. you left out a quite crucial detail, which arch is your system running? 
Ours is running amd64...

> And the following services:
> 
> . OpenVPN tunnel on a tun device (pkgsrc)
> . racoon (5.1 base)
> . named (5.1 base)
> . NFS (5.1 base, rpcbind, mountd, nfsd, rpc.statd, rpc.lockd)
> . rtadvd
> . ISC dhcpd (pkgsrc)
> . Squid (pkgsrc)
> . dansguardian (pkgsrc)
> . nginx (pkgsrc)
> . munin-server and munin-node (pkgsrc)


Not much in common there...

(IPv6 is used, and you seem to have it enabled too...)
From base:
- kerberos
- ntpd
- sshd
From pkgsrc:
- py-denyhosts
- rancid
- conserver8
- tacacs-shrubbery
- pam_krb5
- some perl scripts

I'm not sure if we got any wiser... 
btw. did you have three different IP-networks connect before you started using 
.1Q? (If not, can you try do the same thing without VLAN tagging?)
As you mentioned earlier, another way to go is to upgrade to 6.0_BETA2 to see 
if the problem still exists...

/P