LVS
lvs-users
Google
 
Web LinuxVirtualServer.org

Re: Looking for Simple Instructions

To: "LinuxVirtualServer.org users mailing list." <lvs-users@xxxxxxxxxxxxxxxxxxxxxx>
Subject: Re: Looking for Simple Instructions
From: Roberto Nibali <ratz@xxxxxxxxxxxx>
Date: Fri, 10 Nov 2006 19:01:36 +0100
I've attempted to simplify things on this new attempt (fyi, there is no eth0 on any of the machines):

DIP = 74.52.166.34  bound to eth1
VIP = 74.52.166.35  bound to eth1:35

RS1 = 74.52.166.50  bound to eth1
RS1VIP = 74.52.166.35 bound to lo:35

RS2 = 74.52.166.130 bound to eth1
RS2VIP = 74.52.166.35 bound to lo:35

Ok.

On Director:
  [root@lb1 ~]# sysctl -p
  net.ipv4.conf.lo.arp_ignore = 0
  net.ipv4.conf.lo.arp_announce = 0
  net.ipv4.conf.eth1.arp_ignore = 0
  net.ipv4.conf.eth1.arp_announce = 0
  net.ipv4.conf.all.send_redirects = 1
  net.ipv4.conf.default.send_redirects = 1
  net.ipv4.conf.eth1.send_redirects = 1
  net.ipv4.ip_forward = 0
  net.ipv4.conf.default.rp_filter = 1

You should disable rp_filter.

  net.ipv4.conf.default.accept_source_route = 0

On both RS's:
  net.ipv4.conf.lo.arp_ignore = 1
  net.ipv4.conf.lo.arp_announce = 2
  net.ipv4.conf.eth1.arp_ignore = 1
  net.ipv4.conf.eth1.arp_announce = 2
  net.ipv4.ip_forward = 0
  net.ipv4.conf.default.rp_filter = 1
  net.ipv4.conf.default.accept_source_route = 0

Ok.

Care to show the ipvsadm -L -n output?

[root@lb1 ~]# ipvsadm -L -n
IP Virtual Server version 1.2.0 (size=4096)
Prot LocalAddress:Port Scheduler Flags
  -> RemoteAddress:Port           Forward Weight ActiveConn InActConn
TCP  74.52.166.35:23 rr
  -> 74.52.166.50:23              Route   1      0          0
  -> 74.52.166.130:23             Route   1      0          0

Looks perfect.

The preferred way of dealing with this is by instrumenting arp_{announce,ignore} in the proc-fs.

I've cleared out all the arptables stuff and are trying to use the arp_{announce,ignore} as suggested but I am unsure which interfaces need what setting. The mini-HOWTO isn't too clear on this.

The interface carrying the the VIP, since this is the one we should not send/reply arp probes for the VIP. Only the director needs to reply to arp on the VIP.

Can you tcpdump on the director? Are you sure there's not some filtering of illicit traffic on switch ports on your ISP's side?

Yes. Running "tcpdump -n -i eth1 port 23" on the director shows lots of these when I try and telnet from my home machine:

11:37:45.031014 IP 70.241.143.240.3165 > 74.52.166.35.telnet: S 2050237163:2050237163(0) win 65535 <mss 1452,nop,nop,sackOK>

In earlier days I would have said missing arp handling, yours seems to be ok. So please disable rp_filter and try again. Also check your kernel messages, e.g. the dropped packets from the reverse path filtering go there if log_martians is enabled.

Running "tcpdump -n -i any port 23" on the 2 RS's shows nothing when I try to telnet to the VIP.

Ok, so packets are dropped at the director.

Thanks very much for your assistance.

We're glad to help out, if time permits.

Best regards,
Roberto Nibali, ratz
--
echo '[q]sa[ln0=aln256%Pln256/snlbx]sb3135071790101768542287578439snlbxq' | dc

<Prev in Thread] Current Thread [Next in Thread>