I have an application that does not fail over nicely to another server. I would like to configure my CSM to load balance normally on new connections and keep the sticky so an individual workstation always accesses the same server.
It is my understanding that should Server A fail, eventually all connections will be redirected to Server B. i would like to prevent this from happening. The client would rather see loss of connectivity than have a workstation's connection fail over to the operating Server B.
The only way would be to prevent the server from going down. So, by not configuring probes, as long as the server is able to answer arp request, the traffic should be forwarded it to it and if the service is down, the client will get a failure.
If you also want to cover the server is completely down, you could configure static arp entries on the CSM for those servers or put a routing device between the CSM and the server.
However, I have to question such a config.
Why would you want half the users to go the failed server ?
Are you afraid that one server only would go down with the load ? if so you could use maxconn to protect it.
Are you afraid that when a server goes down all connections get stuck to the remaining server due to the sticky entry ?
If so, you could use a source hash algorithm instead of a sticky group.
All I want to say is that there might be a better solution to your problem than having users stuck to a dead server.
Thanks Gilles, I tend to agree but would not be diligent if I didn't get all the facts. My thoughts are that this cannot be done but have to explore it for the developers.
The main reason that I want users to stick with the same server is that the application itself can fail badly and give script errors if the user swaps servers in the same session. The sticky option will keep the application running correctly normally, but should a server fail completely the session will eventually be swapped and strange things will happen thus confusing the user - stranger & less confusing than losing the app entirely I am told. Who am I to judge?
In the situation of a failed server, I want new sessions & users to go to the good server, so I'm not sure a source hash will be exactly what I need. I would just like any sessions that are active on the failed server to not swap - just fail. If the users starts a new session then that can go to the good server.
Why do you need native HA: The native HA feature allows two Cisco DCNM
appliances to run as active and standby applications, with their
embedded databases synchronized in real time. Therefore, when the active
DCNM is not functioning, the standby DCNM will...
This document will provide screenshots to outline the steps to setup
TACACS+ configuration to ACI and also the configuration required on
Cisco ACS server. Please find the official Cisco guide for configuring
TACACS+ Authentication to ACI:
Is it supported or NOT supported? It's a frequently asked question.
Before APIC, release 2.3(1f), transit routing was not supported within a
single L3Out profile. In APIC, release 2.3(1f) and later, you can
configure transit routing with a single L3Out pr...