Re: [SSI-devel] SSI-1.2.0-FC2 HA-LVS bad ipvs table on failover node

[Aneesh K. <ane...@gm...>]

Hi Roger,


On Thu, 24 Feb 2005 15:11:45 -0500 (EST), Roger Tsang
<op...@bl...> wrote:
> Hi,
> 
> I'm using SSI-1.2.0 Fedora Core 2.  HA-LVS load-balance connection table
> duplication to IPVS directors seems to work fine as expected until the
> following situation occurs.
> 
> I noticed that after failing over the initnode (to node2) and restoring
> the original initnode (node1), the ipvs table is incomplete on the
> original initnode (which now becomes the failover initnode).
> 
> Descending sequence of events
> LVS AOK node1: initnode              - node2: failover initnode
> LVS AOK node1: fails                 - node2: initnode
> LVS BAD node1: failover initnode     - node2: initnode
> 
> I grabbed the ipvsadm output of what I mean by bad ipvs table.  node1 ipvs
> table is missing load-balancing info for node2.
> 
> node1 ICS interface IP 10.117.0.1
> node2 ICS interface IP 10.117.0.2
> 
> (node 1)
> IP Virtual Server version 1.0.10 (size=65536)
> Prot LocalAddress:Port Scheduler Flags
>   -> RemoteAddress:Port           Forward Weight ActiveConn InActConn
> TCP  10.0.0.211:8443 wlc
>   -> 10.117.0.1:8443              Local   3984   0          0
> TCP  10.0.0.211:3456 wlc
>   -> 10.117.0.1:3456              Local   3984   0          0
> TCP  10.0.0.211:143 wlc
>   -> 10.117.0.1:143               Local   3984   0          0
> TCP  10.0.0.211:443 wlc
>   -> 10.117.0.1:443               Local   3984   0          0
> TCP  10.0.0.211:80 wlc
>   -> 10.117.0.1:80                Local   3984   0          2
> TCP  10.0.0.211:22 wlc
>   -> 10.117.0.1:22                Local   3984   0          0
> (node 2)
> IP Virtual Server version 1.0.10 (size=65536)
> Prot LocalAddress:Port Scheduler Flags
>   -> RemoteAddress:Port           Forward Weight ActiveConn InActConn
> TCP  10.0.0.211:8443 wlc
>   -> 10.117.0.2:8443              Local   1471   0          0
>   -> 10.117.0.1:8443              Route   3984   0          0
> TCP  10.0.0.211:3456 wlc
>   -> 10.117.0.2:3456              Local   1471   0          0
>   -> 10.117.0.1:3456              Route   3984   0          0
> TCP  10.0.0.211:143 wlc
>   -> 10.117.0.2:143               Local   1471   0          0
>   -> 10.117.0.1:143               Route   3984   0          0
> TCP  10.0.0.211:443 wlc
>   -> 10.117.0.2:443               Local   1471   0          0
>   -> 10.117.0.1:443               Route   3984   0          0
> TCP  10.0.0.211:80 wlc
>   -> 10.117.0.2:80                Local   1471   0          0
>   -> 10.117.0.1:80                Route   3984   0          0
> TCP  10.0.0.211:22 wlc
>   -> 10.117.0.2:22                Local   1471   0          0
>   -> 10.117.0.1:22                Route   3984   0          0
> 

I remember fixing a simillar bug last time. This is also marked in the
change log

2004/12/8       kvaneesh <ane...@gm...>
        * ha-lvs secondary node boot up was not registering the
        services started during boot up with CVIP. Fix the same


I am right now tracking down the changes to see if i missed something. 

-aneesh