Please can anybody clarify me CSM stickiness behavior when using ip address source or source/destination as sticky criterium.
Problem description: CSM runs version 4.2(2) then I config vserver "WEB" with the sticky group 10
sticky 10 netmask 255.255.255.255 timeout 60
This work find
vserver WEB
virtual 10.9.3.4 tcp www
serverfarm SERVERS_WEB
sticky 60 group 10
replicate csrp sticky
replicate csrp connection
no persistent rebalance
parse-length 4000
inservice
Now we config other vserser "IMP" with sticky group 40
sticky 40 netmask 255.255.255.255 timeout 60
vserver IMP
virtual 10.9.3.8 tcp 0
serverfarm SERVERS_IMP
sticky 60 group 10
replicate csrp sticky
replicate csrp connection
no persistent rebalance
parse-length 4000
inservice
If we connect from the same client, first to vserver WEB, you reach serverfarm "SERVERS_WEB". Now yor try to connect to vserver IMP then you get again serverfarm "SERVERS_WEB" instead serverfarm "SERVERS_IMP".
At the connection time CSM behavior look first for the source ip and sticky hash if mach connect to the same server (never look for the destination (vserver ip)).
To avoid this, I try with sticky groups like this:
sticky 80 netmask 255.255.255.255 address both timeout 60
But as the sticky hash is calculated using and XOR between the source an destination, it very easy to get the same hash for vservers and clients on the same subnet.
Of course now I am using cookies for stickness where I can do it.
Please can anybody tell me is this a normal CSM behavior, my configration are wrong or it´s a CSM firmware bug.
