Willy Tarreau schrieb:
> Hi Corin,
>
> On Tue, Dec 04, 2007 at 09:42:43PM +0100, Corin Langosch wrote:
>
>> hi marc,
>>
>> yes, netfilter with conntrack is running. I saw errors some weeks ago,
>> but i fixed them with:
>>
>> echo "Tuning network settings according to haproxy..."
>> echo 1048576 > /proc/sys/net/ipv4/netfilter/ip_conntrack_max
>> echo 2000 60000 > /proc/sys/net/ipv4/ip_local_port_range
>>
>> from that time on all iptables errors in the syslog were gone. the proxy
>> box is a dual opteron, 0.25 uptime.
>>
>
> It is possible that either your proxy and/or your servers encounter a
> duplex mismatch between their NIC and the switch. This is a very common
> problem. Run ethtool on your server to check that the link is correctly
> reported as Full Duplex. If the switch is forced to 100Full with a gigabit
> NIC in the server set to auto, you generally end up with 100 Half on the
> server. That alone could explain all what you are observing.
>
>
>> by the way: which timeout values uses haproxy for the httpchk option?
>>
>
> Since only one check can be running at a time on a server, the "inter"
> parameter serves both for the check interval and the check timeout.
>
> If you increase it above 3 seconds and you notice an important drop in
> the number of check failures, it would most likely indicate a network
> problem causing retransmissions to happen.
>
> Regards,
> Willy
Thank's for the information. I grabbed ethtool, the output looks good:
Settings for eth0:
Supported ports: [ TP ]
Supported link modes: 10baseT/Half 10baseT/Full
100baseT/Half 100baseT/Full
1000baseT/Full
Supports auto-negotiation: Yes
Advertised link modes: 10baseT/Half 10baseT/Full
100baseT/Half 100baseT/Full
1000baseT/Full
Advertised auto-negotiation: Yes
Speed: 1000Mb/s
Duplex: Full
Port: Twisted Pair
PHYAD: 0
Transceiver: internal
Auto-negotiation: on
Supports Wake-on: umbg
Wake-on: g
Current message level: 0x00000007 (7)
Link detected: yes
ifconfig does not show any errors or overflows.
is there a way to tell haproxy to output a detailed report on failed requests?
changing inter from 1000 to 5000 did not really change anything. I now put it back to 1000 and upgraded the kernel, maybe (hopefully) it was a kernel bug. i'll keep the list informed.
regards,
corin
Received on 2007/12/05 09:25
This archive was generated by hypermail 2.2.0 : 2007/12/05 09:30 CET