I replaced my old piranha powered load balancer with a new keepalived powered one. keepalived keeled over and died.

What's up? I have no idea.

The situation:

I have a heavily loaded site that's fronted by piranha on an old i686 128 M RAM. I thought that it's a good time i replaced that with a keepalived installation. I've used keepalived on another site with no problems at all.

So, replaced piranha, I did. After a few minutes, kepalived inhibited all my real servers. What the? Fiddled with the connect_timeout, nb_get_retry, and delay_before_entry, still the same.

Here's the value that I tried:
-30, 10, 1
-120, 50, 1
-60, 10, 1

I have no problems with piranha. the ipvsadm with piranha:

IP Virtual Server version 0.8.2 (size=65536)

Prot LocalAddress:Port Scheduler Flags

  -> RemoteAddress:Port             Forward Weight ActiveConn InActConn

TCP  20x.x.x.x:http wlc

  -> 19x.x.x.x:http              Masq    1      58         5707

TCP  20x.x.x.x:http wlc

  -> i:http      Masq    1      0          0

TCP  20x.x.x.x:http wlc persistent 1

  -> s4:http          Masq    3000   210        1

  -> s9:http          Masq    5500   357        1303

  -> s10:http         Masq    5500   402        1058

  -> s2:http          Masq    5500   317        3006

  -> s11:http         Masq    5500   319        3619

  -> s7:http          Masq    4000   337        1437

  -> s8:http          Masq    6500   276        8644

Now, keepalived works with another site, as I've mentioned. The only difference is the load.

ipvsadm of the site (the one with less load):

IP Virtual Server version 1.2.0 (size=4096)

Prot LocalAddress:Port Scheduler Flags

  -> RemoteAddress:Port           Forward Weight ActiveConn InActConn

TCP  20x.x.x.x:www wlc

  -> s5:www         Masq    1      5          1212

  -> s4:www         Masq    1      6          1139

  -> s3:www         Masq    1      7          946

So, reverted to piranha, I did. :(

Still trying to make sense of this.