Forum Discussion
question for load balancing
hi~
I have some problems configuring load balancing.
Device: BIG-IP 10250v
Environment: server1, server2, server3
Each server has 10 identical services, each using a different port.
service is web site(http)
server1 (xxx.xxx.xxx.011): ins0 8080, ins1 8081 ... ins9 8089,
server2 (xxx.xxx.xxx.012): ins0 8080, ins1 8081 ... ins9 8089,
server3 (xxx.xxx.xxx.013): ins0 8080, ins1 8081 ... ins9 8089,
L4: xxx.xxx.xxx.010: 8080 ... 8089
least connection load balancing
Load balancing testing with http request detected anomalies.
If you send 1 million http requests continuously for a certain time
The number of operations per instance group is similar,
but the number of operations per server is significantly different.
ins0 (8080) group: 11%
ins1 (8081) group: 8%
...
ins9 (8089) group: 10%
The difference in the number of processes between each instance group is similar to 1-2%.
server1: 34%
server2: 41%
server3: 10%
(instance group is Horizontal perspective, server group is Vertical perspective)
server1 server2 server3 sum
ins0 19 79 9 11%
ins1 x x x 8%
ins2 x x x 10%
... x x x ..%
ins9 x x x 11%
sum 34% 41% 10%
However, the number of processes per server is very different.
What's the big difference?
- rob_carr
Cirrocumulus
The number of operations per instance group is similar, but the number of operations per server is significantly different.
What does the term 'operations' mean in this sentence? Are you referring to TCP connections, HTTP requests or CPU cycles (for example)?
- numi_372246
Nimbostratus
'operation' is HTTP requests in was access log.
- rob_carr
Cirrocumulus
How was your test setup? Did you have a pool for each service, with one instance from each of your three physical servers?
- numi_372246
Nimbostratus
each physical server has ten instance. we have 3 physical server.
We solved the problem with multiple tests.
The problem was the session retention time. The four test PCs consistently generated http requests with a fixed IP address. However, with the session retention time of 300 seconds, the request was continuously transmitted only to the server that processed the initial request, and the load balancing could not be performed evenly.
With continuous continuous requests, the session retention time was constantly maintained at 300 seconds This prevents load balancing to other servers until the server processing the request is stopped.
We solved the problem to equalize load balancing by releasing session time(setting 0).
thanks rob_carr
Recent Discussions
Related Content
* Getting Started on DevCentral
* Community Guidelines
* Community Terms of Use / EULA
* Community Ranking Explained
* Community Resources
* Contact the DevCentral Team
* Update MFA on account.f5.com