Help me understand Load Balancing

Question

Good afternoon everyone,&nbsp;I am hoping someone can help me understand the difference between something that is "failing a healthcheck" in an F5 and something that is "Forced Offline" and how the load balancer would react to both.At my company I notice that if I have a server failing the healthcheck in the load balancer, that load balancer will still send requests to that server experiencing issues. But if I force that server offline manually, then the load balancer respects that the server is down and doesn't send it any requests until we bring it back up manually. Is this the expected behavior from an F5 load balancer? Or does it depend on the version of the device in question or the software?&nbsp;According to the manager that runs this system, they are telling me this is how it is and that the load balancer isn't "smart" enough to know unless we manually force it offline. Does this pass the sniff test or they being misleading? To me this sounds misleading at best. Because what is the point of having an active health check if the load balancer is still going to send requests to servers that fail the health check?I am just trying to educate myself on this and since this is not my area of expertise. I would think a load balancer should be smart enough OOB to handle functionality like this. But I also want to make sure I am not "inventing" functionality that may not be there or is supported through a different license. Any type of info would be appreciated and thank you in advance for anyone who takes the time to read and reply to this post!&nbsp;Respectfully,&nbsp;Brian Jones

aswin_mk · Answer

HelloFor education you can use F5 article and devCentral youtub vedios. Just search your query in in Google and you will get article relating to that. Also so many education vedios are available in YouTube devCentral page.&nbsp;For eg : https://techdocs.f5.com/kb/en-us/products/big-ip_ltm/manuals/product/ltm-monitors-reference-12-0-0/2.html

mohamed_ahmed_kansoh · Answer

Hi @bjones2480,&nbsp;
Review the timers or the intervals that you set in this Active Health monitor if it's in the default values ( Interval 5 seconds &amp; Timeout on 16 Seconds ) Bigip will mark this Pool member as an offline "Red Rhombus" after 16 seconds then it will not send any further requests.
But within 16 second it will use this pool member in load balancing decisions.&nbsp;
&nbsp;
you can lower the interval and timeout values, so Bigip will mark it offline quickly and prevent any requests from passing to this pool member.

bjones2480 · Answer

Mohamed,&nbsp;Thank you for this information and we are in fact using the defaults.&nbsp;If I understand you right. 5 seconds is the duration between the health checks and 16 seconds is the time the server has to respond before being marked as failing the health check?However what we are seeing is the LTM sending requests to these servers failing the health check. Is there an instance or a config that would allow this to happen in BigIP?&nbsp;&nbsp;&nbsp;

mohamed_ahmed_kansoh · Answer

Hi,&nbsp;Do you mean Bigip sends requests to a down pool member even after the monitor marks it down ??!!&nbsp;I doubt that happens, Bigip will send requests till the timeout value reached ~"16" seconds but after that the assigned health monitor will mark this pool member as down and stop sending traffic to it.
&nbsp;
Take a packet capture for a minute to validate that&nbsp;or reset pool member stats after being down and see packet in/out to confirm this.

bjones2480 · Answer

Mohamed,&nbsp;I verified this functionality after noticing proxy errors from one of my systems intermittently. When I investigated the load balancing pool members I noticed one was failing a health check. As long as that server remained red in the load balancer (failing health check) I would get those intermittent proxy errors. As soon as I forced that failing pool member offline the intermittent errors would disappear.&nbsp;I reached out to the team that manages our load balancing and they told me the following:"Existing connections are the connections which were going to the server when it was up, before failing healthcheck. Those connections would continue to connect to failed server until they are terminated by the client itself or they are forced to terminate manually by us. Changing the load balancing method will not help as existing connections will still continue to connect to failed server"Is this the timing issue you were speaking of? What existing connections are they talking about if the system is configured for round robin load balancing, wouldn't each request be a new connection?"I will have to agree with you load balancers not smart enough, unfortunately that’s what we have noticed over the last few years and the only way to get the fresh connections initiate is to either force offline the server or kill the connections that’s connected to the failed server." This comment seems to fly in the face of everything I have read from BigIP and what you are telling me. So I am trying to understand if I am not conveying my question properly to them or not. While both of these comments make me question the purpose of a health check for a load balancer in our org if it isn't affecting how the system balances the load.&nbsp;&nbsp;I am just trying to educate myself enough to take this concern to them and speak to it coherently. So I appreciate you taking the time to answer these questions.&nbsp;

Forum Discussion

Help me understand Load Balancing

Recent Discussions

201 Recommendations for Study

GTM pool is OFFLINE even if pool members are UNKNOWN

URL redirect

Vertica DB Monitoring Failure post upgrade 10.x to 23.x

SNMP Health Monitors

Related Content

Understanding iApps

Understanding IPSec IKEv2 negotiation on Wireshark

Understanding IPSec IKEv1 negotiation on Wireshark

Understanding Modern Application Architecture - Part 1

Active/Active load balancing examples with F5 BIG-IP and Azure load balancer