The pool members up/down for no reason

Either you have a network issue which is causing your monitors to fail, OR your servers are having problems you aren't aware of OR the monitors themselves need to be tuned to be more appropriate to your specific servers.

A monitor will fail if;

• the server does not respond
• the server does not respond in time
• the server does not respond with the expected response

You need to perform a tcpdump to ascertain which one (or more) of these is causing your issue. The following will display hex/ascii of healthcheck traffic on the screen;-

tcpdump -i0.0:nnn -s0 -XXX -vv host x.x.x.x and port yy

where x.x.x.x is the non-floating self-ip of your LTM, and yy is the port used by your pool.

Or if that causes too much traffic you can output to file and view in Wireshark;-

tcpdump -i0.0:nnn -s0 -w /var/tmp/health.cap host x.x.x.x and port yy

What kind of monitors are you using?

Hi, This is known issue of servers(Windows/Unix, etc). I guess server agent(Services) are failing. Please restart server one by one. Problem will resolved.

Regards, Samir

We've also seen this happen with garbage collection issues on the servers, although there has been impact on server performance at these times.

its happened due to memory buffer overflow in server. You just reboot server on role on basis(only option).

To start with, you can try to change the monitor from L7(assuming you are on HTTP/HTTPS) to L3 (ICMP), so that you can be sure about whether the problem lies in network connectivity OR the service itself. You can use ICMP and HTTP both monitors at the same time. Just select "at least one" option for getting logical OR between the two monitors.

Hey,

This problem rises if you are using iCMP as helth monitor. Sometimes echo reply-request RT cause this problem. You disable the node level health check and enable custom health monitor (say http/https/tcp/udp) in the pool level. It will work.

Let me know if above solution work for you.

-Jinshu

Just ran into something similar during LTM deployment. One of self IPs had a duplicate on the network.