Localdir 416 real servers not failing..

[mbellears]

We have a localdirector sitting in front of two real servers(IIS).

Load balancing works fine with both backend servers connected, and show real produces:

# show real

Real Machines:

No Answer TCP Reset DataIn

Machine Connect State Thresh Reassigns Reassigns Conns

server2:0:0:tcp 2 IS 8 0 0 0

server1:0:0:tcp 0 IS 8 0 0 0

But if one backend server is disconnected, show real does not change (No OOS, or TESTING under STATE), and nothing is displayed in syslog?

ping server1

real_server_ip_1 NO response received -- 1000ms

real_server_ip_1 NO response received -- 1000ms

real_server_ip_1 NO response received -- 1000ms

show real

Real Machines:

No Answer TCP Reset DataIn

Machine Connect State Thresh Reassigns Reassigns Conns

server2:0:0:tcp 2 IS 8 0 0 0

server1:0:0:tcp 1 IS 8 0 0 0

Is this normal?

Minimal config, just for testing:

virtual virt_ip:0:0:tcp is

real real_server_ip_1:0:0:tcp is

real real_server_ip_2:0:0:tcp is

name real_server_ip_1 server1

name real_server_ip_2 server2

name virt_ip domain

bind virt_ip:0:0:tcp real_server_ip_1:0:0:tcp

bind virt_ip:0:0:tcp real_server_ip_2:0:0:tcp

Regards,

MB

[mbellears]

Also - Upon failure (Eg One of the Backend Servers are disconected), there is a 30 second window were website access is unavailable - Is this normal behaviour?

[Gilles Dufour]

the LD does not failover a service as soon as you disconnect it.

It will still try to use the service and if no response is seen from the server it will end up marking it down.

But for that, you need new connections attempt first.

That's why you see the website as down for 30 seconds.

To detect the real server down, you can configure probes. This is traffic generated by the LD which will follow the same rules as above.

Hope this help.

Gilles.

[mbellears]

Thanks to all who replied - I setup probes yesterday, and they work like a charm!

My testing obviously did not replicate 'real-world' connection attempts - hence the 30 sec. delay.

Regards,

MB

[jkanclirz]

Depending on your version, here is a good document how servers are failed and brought back on LD.

http://www.cisco.com/univercd/cc/td/doc/product/iaabu/localdir/ld33rns/ld334con/ld3_ch01.htm#xtocid275378

"

The reassign command controls how many times a connection synchronization (TCP SYN) packet from a requesting client is sent to a nonresponsive server before it is reassigned to another server. The default is three TCP SYN packets. After the third packet receives no response or a TCP RST from the server, the fourth packet is sent to another server.

Each reassign process increments the reassign tally by one. When the tally reaches the threshold value, the server is considered failed. With a default threshold value of 8, the reassign process will happen eight times before the server is considered failed. "

In other words, the LD doesn't ping and check the server if its up or down, it takes the client(end-user). Depending on your site, if you have a very slow active site... it could take that much more time for LD to fail the down server.

-jan