RE: connection sync at failover, email, and using only basic IPmgmt

["Richard Pickett" <Richard.Pickett@xxxxxxxxxxxxxxxxxxx>]

> On Mon, 2006-02-13 at 14:02 -0600, Richard Pickett wrote:
> > Since I'm only using keepalived the current connections aren't
carried
> > over to the backup at sync. I'm running Fedora 4 with the 2.6.11
kernel
> > (can easily upgrade to 2.6.15). But I don't see any IPVS syncd
patches
> > for this kernel. Is it included? And if yes, how do I configure it
to
> > sync up between the master and backup?
> 
> Using keepalived you will need the "lvs_sync_daemon_interface"
> configuration option for a given VRRP instance. If you have the same
on
> both your MASTER and BACKUP, the BACKUP will sync state correctly and
> connection states should carry over on failover.

I've done that. It doesn't seem to sync the data at all. We have a
x-over cable connecting the two boxes on eth2 w/ a valid network setup
on the nics and ping on that connection works fine.

When we fire up keepalived on both boxes and run tcpdump -I eth2 we se
*no* traffic no matter how many connections come on or how long the
connection runs (some are downloads that last over a minute).

My master conf is below. Any ideas?

> On failback, however, you may need to make use of the preempt_delay
> option to give the machine which will become MASTER time to populate
> its' state table.

I tried that too. I do the preempt_delay, nopreempt, etc. It doesn't
work. As soon as the master box comes up he owns all the IPs. I've ran
this in many different combinations (maybe not the right one yet), where
both boxes are labeled BACKUP, where both have preempt* instructions,
where only one does and not the other, switched that around, etc... No
matter how I do it, as soon as the box with the higher priority comes up
he takes over.

Maybe the 2.6.11 kernel doesn't have the right stuff in it?

Here's my master conf, I change "MASTER" to BACKUP, a lower priority on
each, and the lvs_id to *02 for the backup config. I'm only using simple
TCP_CONNECT for now, didn't want to complicate things by actually
pulling a web page until I get some of these other issues resolved.

Thanks for all you guy's great help!

global_defs
{
   lvs_id LVS_HA_01
}

vrrp_sync_group VG1
{
   group
   {
      VI_1
      VI_GATEWAY
   }
}
vrrp_instance VI_1
{
   state MASTER
   interface eth0

   lvs_sync_daemon_inteface eth2
   virtual_router_id 51
   priority 150
   advert_int 1
   authentication
   {
      auth_type PASS
      auth_pass xO_pIa6ie
   }

   virtual_ipaddress
   {
      172.16.0.4
   }
}
vrrp_instance VI_GATEWAY
{
   state MASTER
   interface eth1
   lvs_sync_daemon_inteface eth2
   virtual_router_id 52
   priority 150
   advert_int 1
   authentication
   {
      auth_type PASS
      auth_pass xO_pIa6ie
   }
   virtual_ipaddress
   {
      172.16.2.1
   }
}
virtual_server 172.16.0.4 80
{
   delay_loop 6
   lb_algo rr
   lb_kind NAT
   nat_mask 255.255.255.0
   protocol TCP
   real_server 172.16.2.4 80
   {
      weight 1
      TCP_CHECK
      {
         connect_timeout 3
         connect_port 80
      }
   }
   real_server 172.16.2.5 80
   {
      weight 1
      TCP_CHECK
      {
         connect_timeout 3
         connect_port 80
      }
   }
}