|
On Sat, Oct 25, 2008 at 08:43:25PM +0100, James Wilkins wrote:
> Hello,
>
> Apologies if this has been covered before, i've not been able to find
> information on it. Any input/suggestions on this issue appreciated.
>
> We are running a 2-node HA Cluster utilising ldirectord - at present
> we are running multiple instances of ldirectord - each instance of
> ldirectord handling the checks for one VIP.
>
> Sporadically we are seeing an issue where rogue nodes from another VIP
> are ending up inside the first VIP (shown by ipvsadm). To explain
> this better;
>
> Normal config:
>
> VIP1:
>
> TCP 1.2.3.4:443 rr persistent 600
> -> 192.168.0.1:443 Masq 1 0 0
> -> 192.168.0.2:443 Masq 1 0 0
> -> 192.168.0.3:443 Masq 1 0 0
>
> VIP2:
>
> TCP 1.2.3.5:443 rr persistent 600
> -> 192.168.0.1:444 Masq 1 0 0
> -> 192.168.0.2:444 Masq 1 0 0
> -> 192.168.0.3:444 Masq 1 0 0
>
>
> Problem occured - notice the addition of the two entries with port 445
> from VIP2 ending up in VIP1.
>
> VIP 1:
>
> TCP 1.2.3.4:443 rr persistent 600
> -> 192.168.0.1:443 Masq 1 0 0
> -> 192.168.0.2:443 Masq 1 0 0
> -> 192.168.0.3:443 Masq 1 0 0
> -> 192.168.0.1:444 Masq 1 0 0
> -> 192.168.0.2:444 Masq 1 0 0
>
> VIP 2:
>
> TCP 1.2.3.5:443 rr persistent 600
> -> 192.168.0.1:444 Masq 1 0 0
> -> 192.168.0.2:444 Masq 1 0 0
> -> 192.168.0.3:444 Masq 1 0 0
>
> Hoping someone can advise if they've seen this before?
Hi James,
that looks very curious and smells a lot like a bug in ldirectord.
Which version are your running?
--
Simon Horman
VA Linux Systems Japan K.K., Sydney, Australia Satellite Office
H: www.vergenet.net/~horms/ W: www.valinux.co.jp/en
|
