Forum Discussion

Teamnetzbetrieb's avatar
Teamnetzbetrieb
Icon for Nimbostratus rankNimbostratus
Jan 23, 2020

Dynamic Ratio for Proxy Server

Hi,

we would like two proxy servers fw-proxy-01 and fw-proxy-09 via the F5 balancing, whereby the load distribution should take place via dynamic ratio.

For this we have created a dca basic monitor which queries the current load via snmp with the oid .1.3.6.1.4.1.2021.10.1.3.1. This monitor was assigned to the nodes.

In the pool, the load balancing method is configured to "Dynamic Ratio (member). The configured ratio of each node is 1.

But although fw-proxy-01 shows a load of 0.25 and fw-proxy-09 a load of 0.01 (snmp dca logging activated) the sessions are still routed to fw-proxy-01:

 

========== /usr/share/monitors/SNMPDCABASE_monitor ==========

Node Name: /Proxyfarm/fw-proxy-01

Node Address: ::ffff:x.x.x.1

SNMP version: v2c, community = <sensitive data redacted>, port: 161

Monitor: snmp_dca_base.

 

User OIDs:

        .1.3.6.1.4.1.2021.10.1.3.1 - 0.250000

 

i = 0, util = 0.250000, threshold=90.000000, coeff=1.000000

 

Main:Weight =10 (weight1 = 9.936243, weight2 = 0.000000)

 

========== /usr/share/monitors/SNMPDCABASE_monitor ==========

Node Name: /Proxyfarm/fw-proxy-09

Node Address: ::ffff:x.x.x.9

SNMP version: v2c, community = <sensitive data redacted>, port: 161

Monitor: snmp_dca_base.

 

User OIDs:

        .1.3.6.1.4.1.2021.10.1.3.1 - 0.010000

 

i = 0, util = 0.010000, threshold=90.000000, coeff=1.000000

 

Main:Weight =10 (weight1 = 9.997442, weight2 = 0.000000)

 

The tmsh command "list ltm node dynamic-ratio" show me always a dynamic ratio osf 10 for both nodes:

 

ltm node fw-proxy-01 {

   dynamic-ratio 10

}

ltm node fw-proxy-09 {

   dynamic-ratio 10

}

 

Does anyone know, how we can create a dca basic monitor for this scenario? Is it necessairy calculate any values with coefficient and / or thresholds?

And if so, how can we do that?

 

Regards Michael.

 

 

3 Replies

  • Hi

     

    It seems you are in a situation where this is not supposed to work

     

    https://support.f5.com/csp/article/K9125

     

    Note: Currently the Dynamic Ratio (member) option is only available when used with a Server/Application State Protocol (SASP) type monitor. When other monitor types are in use, choosing Dynamic Ratio (member) load balancing will cause the BIG-IP system to distribute traffic evenly to all pool members.

     

    So I guess you will have to change things a bit...

     

    I have to admin I am not very familiar with this, but hopefully someone else arount will :) ?

     

    Yoann

  • Hi Yoann,

    thank you for your comment. But, sorry I forget, we use firmware 14.1.2, so I think this article is more applicable for us.:

    https://techdocs.f5.com/en-us/bigip-14-1-0/big-ip-local-traffic-management-basics-14-1-0/dynamic-ratio-load-balancing.html

    We use the snmpd installed on the server, so that the UC Davis SNMP agent is used. And as you can see, the weights are calculated - maybe the server load is not high enough.

     

    Your article is for firmware up to 13.x.

  • Hi

     

    Indeed. :)

     

    It is not clear whether the calculation algorithm is changed, but, there may be other issue. As you are basing your decision based on only CPU right ?

     

    I can't see much detail on the calculation formula in v14 (don't know if it changed from the one mentionned in the KB article I sent...

     

    I think this may still be applicable :

     

    The BIG-IP system rounds the values to the nearest whole number.

     

    Your values would always be rounded to 10. Try to play with the coefficient, or add memory/disk values to enrich the wheight value.

     

    Yoann