Please Read: Significant Update Planned, Migrating Forum Software This Month

See more
See less

Switch Redundancy - STP

  • Filter
  • Time
  • Show
Clear All
new posts

  • Switch Redundancy - STP

    Greetings all - we've recently upgraded our VI setup and am now looking into improving our uptime in case of other hardware failures - particularly our networking topology.

    In one rack (RACK-1) we have our ESX hosts, iSCSI SAN and remaining physical servers which are connected to 2 Procurve 2910al-48G's (switches A and B).

    We have a number of VLAN's configured on each (production, storage, vmotion, test, etc) and a single uplink between them. All servers and storage are patched to both switches so that if either switch fails we're still good.

    RACK-2 houses our other switches and patching to clients:
    - Switch C: 1 x Procurve 2848 (GB) - a few physical servers, router, IT staff, Video Conf, various mgmt clients, ports for connnecting to above VLAN's
    - Switch D,E,F: 3 x Procurve 2650 - staff PC's, printers, etc

    Switch C is uplinked to D,E,F
    Switch C is uplinked to A and configured with the same VLANS as A & B.

    So while the machines on RACK-1 can survive a switch failure of A or B, everything on RACK-2 will lose connectivity if A fails given its the only uplink.

    Ideally we would like to uplink switch B to one of the switches in Rack 2 to remove this point of failure however this brings loops into play.

    I've been reading up on STP and have a general understanding of how it is meant to 'block' the redundant link until its needed, however get a little confused with how/if VLAN's come into play, RSTP, MSTP, HP's 'meshing' technology, etc.

    Whats the best topology to add in this redundancy? Should I just enable STP on all and add the second uplink from B to Switch C, or uplink B to one of the other switches to form a 'ring'?

    Some other info:
    - All switches appear to support STP (currently disabled)
    - the 2650's (D,E,F) are 10/100 switches however have 2 1000 ports (no VLANS currently configured just accepts untagged production traffic)
    - the uplink between switches A & B is via a single 1000 connection
    - there's (currently) no routing required between VLAN's.

    Thanks in advance,
    Last edited by bunce; 27th March 2010, 15:55.