Why might I do this you ask? Basically for diagnostic purposes. Obviously I have monitors, but I need my own source of truth to know whether or not services on my pool members are up.
I am using LTM to load balance syslog. I require redundancy at the pool member level, and just one server isn't enough to handle all the traffic.
I run a simple bash script which generates syslog messages and shoots it at all the ports and protocols open for syslog on the box itself, and those that sit on the load balancer. However, when those diagnostic messages hit the load balancer, they never arrive at the other pool members. Only the pool member that originates the traffic gets traffic from the load balancer. Node statistics show traffic going to all pool members, but the other pool members don't see it.
Virtual server is stateless UDP. Datagram LB enabled. No SNAT or AutoMap (because I must preserve the true source IP of the originating host, because sometimes that's all I have to filter on). If I SNAT, it works. So I suspect this is just me not understanding layer 2/3/4 well enough. Same behavior when using TCP syslog via the F5 as well.
No comments:
Post a Comment