Problem solved .. but unable to use VIP for all ports using FW mark

[Ranga Nathan <kairanga@xxxxxxx>]

I traced it down to my ldirectord configuration. I assumed that the VIP 
is available for all ports. But the configuration opens only port 80 and 
I was testing with ssh connection. I changed the port to 22, and it 
worked fine.
Now I really want to cluster for all ports for a given VIP. According to 
the instructions I should use a firewall mark for this. So I now have:
Chain PREROUTING (policy ACCEPT)
target     prot opt source               destination
          all  --  172.21.113.89        anywhere            MARK match 0x64

and in ipvsadm.rules:
-A  -f 100 -s rr -p
-a -f 100  -r 172.21.113.87  -g -w 1
-a -f 100  -r 172.21.113.88  -g -w 1

and the ldirectord configuration is:
checktimeout=20
checkinterval=5
autoreload=yes
quiescent=no
logfile="info"
virtual=100
   real=127.0.0.1:0 gate 1 ".healthcheck.html" "OKAY"
   real=172.21.113.87:0 gate 1 ".healthcheck.html" "OKAY"
   real=172.21.113.88:0 gate 1 ".healthcheck.html" "OKAY"
   service=http
   checkport=80
   protocol=fwm
   scheduler=wrr
   checktype=negotiate
   fallback=127.0.0.1

But all access to the 172.21.113.89 go to the load director!

Ranga Nathan wrote:

> When a real server goes down ( I forced reboot), ldirectord does not 
> remove the entry from ipvsadm table. As a result a reconnection to the 
> common IP from the same client fails. It tries to go to the same 
> client. I am using roundrobin. I expect ldirectord to remove an entry 
> as soon as it loses the http link. From then on any new connection 
> should use the remaining real servers.
> This is a shot of my ipvsadm  watch ...
> Every 2s: ipvsadm -L -n                                 Fri Nov 18 
> 23:05:32 2005
> IP Virtual Server version 1.2.0 (size=4096)
> Prot LocalAddress:Port Scheduler Flags
>  -> RemoteAddress:Port           Forward Weight ActiveConn InActConn
> TCP  172.21.113.89:80 wrr
>  -> 172.21.113.88:80             Route   1      0          0
>  -> 172.21.113.87:80             Route   1      0          0
> TCP  172.21.113.89:0 rr persistent 360
>  -> 172.21.113.88:0              Route   1      0          0
>  -> 172.21.113.87:0              Route   1      1          0
>
>
> Here is the config for ldirecgtord ( I created based on an example)
>
> checktimeout=20
> checkinterval=5
> autoreload=yes
> quiescent=no
> logfile="info"
> virtual=172.21.113.89:80
>    real=127.0.0.1:80 gate 1 ".healthcheck.html" "OKAY"
>    real=172.21.113.87:80 gate 1 ".healthcheck.html" "OKAY"
>    real=172.21.113.88:80 gate 1 ".healthcheck.html" "OKAY"
>    service=http
>    checkport=80
>    protocol=tcp
>    scheduler=wrr
>    checktype=negotiate
>    fallback=127.0.0.1
>
> Here is the network config for the load director:
> eth0      Link encap:Ethernet  HWaddr 00:08:74:32:4B:73
>          inet addr:172.21.112.1  Bcast:172.21.115.255  Mask:255.255.252.0
>          inet6 addr: fe80::208:74ff:fe32:4b73/64 Scope:Link
>          UP BROADCAST NOTRAILERS RUNNING MULTICAST  MTU:1500  Metric:1
>          RX packets:3542 errors:0 dropped:0 overruns:0 frame:0
>          TX packets:2250 errors:0 dropped:0 overruns:0 carrier:0
>          collisions:2 txqueuelen:1000
>          RX bytes:449504 (438.9 Kb)  TX bytes:261717 (255.5 Kb)
>
> eth0:0    Link encap:Ethernet  HWaddr 00:08:74:32:4B:73
>          inet addr:172.21.113.89  Bcast:172.21.115.255  
> Mask:255.255.252.0
>          UP BROADCAST NOTRAILERS RUNNING MULTICAST  MTU:1500  Metric:1
>
> _______________________________________________
> LinuxVirtualServer.org mailing list - lvs-users@xxxxxxxxxxxxxxxxxxxxxx
> Send requests to lvs-users-request@xxxxxxxxxxxxxxxxxxxxxx
> or go to http://www.in-addr.de/mailman/listinfo/lvs-users