Azure load balancer rule not working

1
votes

I'm new to Azure Load balancing.. so I'm trying to add 3 VM's (v2) to a load balancer(v2) and configured this as follows:

  1. Added 3 VM's with an availability set. Created Load Balancer and used the Availability set.
  2. Added inbound firewall rule for the TCP port on all 3 VM's
  3. Configured inbound NAT rules so that the client can hit for a specific VM at a particular port internally -This worked fine!

Now I need to test a scenario when I need clients to send data to a specific TCP port which should be distributed to the 3 VM's and I did the following steps:

  1. Removed the previously created NAT Rules.
  2. Added the 3 VM's under a single back-end pool.
  3. Added Probes to non http TCP port.
  4. Created Load Balancer rule where I have defined the incoming port and the outgoing TCP port which is different.

This does not seem to be working, any idea why ?

I have 3 VM's configured in a backend pool and need the public network to hit the LB on a custom tcp port, which will inturn hit the VM's randomly configured in the pool. The probes are configured to connect the custom TCP port. The three VM's make use of a common NSG that has inbound rules to allow connections from the incoming custom TCP port.

The port at which the service (td-agent/Fluentd) runs is 24224 on the 3 VM's Current issues:

1) Unable to telnet to the custom TCP port from the LB's public IP. But this works if I create an inbound NAT rule for that port (after deleting the LB rule)that hits to a particular VM, which is not my requirement.

2) Diagnostics for the LB are not writing to the storage as configured to check if the probes are working.

3) No tool or command to check or simulate how the probe connects, if there is one please let me know.

This has also been posted here: https://social.msdn.microsoft.com/Forums/azure/en-US/fca163f9-0d41-4a42-8278-0caaed91b3ff/azure-load-balancer-rule-not-working

1
azureload-balancingfluent

1 Answers

0
votes

Since I was using the fluentd service, I had followed some pre-installation steps which seems to have caused the issue for the Azure load balancer under 'Optimize Network Kernel Parameters' http://docs.fluentd.org/articles/before-install The attribute net.ipv4.tcp_tw_recycle is the one that had caused issues to the LB. A discussion on that can be found here in this link.

On removing this config I was able to see acknowledgments from the Azure LB IP using the command on the VM's :

tcpdump -i eth0 src 168.63.129.16 and port 24224

But, the Diagnostics are still not working as that feature is still in Preview mode.