Ad Widget

Collapse

Using F5 LTM and GTM for HA proxied environment

Collapse
X
 
  • Time
  • Show
Clear All
new posts
  • BSeveryday
    Junior Member
    • Feb 2017
    • 1

    #1

    Using F5 LTM and GTM for HA proxied environment

    I am designing the architecture for a moderately large installation that will include hosts in multiple zones across geographically dispersed datacenters.

    In the whole company we standardized on F5's global traffic manager (GTM) and local traffic manager (LTM) for HA and load balancing of services.

    While we could have all hosts directly contact one server in the home DC I think it is better to use proxies.

    To optimize performance while retaining reliability I'd like to use the following setup for proxies to hosts using the topology balancing for best performance.

    Hosts(100s+) -> [Proxy VIP on GTM] -> LTMs(3+) -> Proxy servers

    Server to proxies could go direct but would prefer to use LTMs for redundancy.

    Does it matter which proxy a host talks to for active checks sending data to the server?

    I realize that many use software based HA and monitors but for larger firms load balancers are a common pattern.

    Any thoughts or advice regarding F5 load balancers for HA / performance
  • kloczek
    Senior Member
    • Jun 2006
    • 1771

    #2
    1) phrases "I am designing the architecture for a moderately large installation" and "hosts directly contact one server" are in contradiction (bigger installation than is more reasons to monitor nothing except zabbix server over server)
    2) "I am designing the architecture for a moderately" and "Hosts(100s+)" is in in a kind of contradiction (atm +100 hosts it may be something between small to mid size installation)
    3) Architecting something and using "I think it is" phrase is usually in contradiction (architecting is about designing something using set of conditions which needs to be fulfilled in final setup .. about which here we know nothing)
    4) you may or may not need use load balancers (like F5 LTMs) depends on level HA you need to guarantee in your project on network layer. Such network layer design aspect depends as well on availability of redundant physical network paths. Zabbix stack may not need 99.99% availability on network layer as using proxies in your setup may cause that such installation in exact points will be be fault tolerant.
    http://uk.linkedin.com/pub/tomasz-k%...zko/6/940/430/
    https://kloczek.wordpress.com/
    zapish - Zabbix API SHell binding https://github.com/kloczek/zapish
    My zabbix templates https://github.com/kloczek/zabbix-templates

    Comment

    Working...