Conference ssdevo::hsj40_product

1224.0. "HSJ40 cache shows as UNKNOWN CACHE TYPE" by CSC32::GULDEN () Sun Mar 23 1997 11:56

       My name is Wes Gulden. I work aat the CSC in Colorado. I worked with a
    MCS Engineer Sunday who had a a failed HSJ40 in a dual-redundant 
    configuration. Each was running 32MB write back cache with HSOF V2.7J-1. 
    The reset light was on solid. He went out to site with two HSJ's
    (70-30097-01 and 70-32193-01). He replaced the bad controller with the
    70-30097-01 spare. He was able to get both controllers back into a dual
    redundant configuration. The new controller would show invalid cache
    that was cleared by doing the command
    CLEAR_ERRORS INVALID_CACHE THIS_CONTROLLER NODESTROY_UNFLUSHED_DATA
    However after doing this, the good controller shows it has 
    UNKNOWN CACHE TYPE.
    
      We attempted to clear it by making the failing cache be a read cache by
    moving the jumper and rebooting then setting it back to write cache. 
    This made no difference. The controller that remains up (not swapped
    out or removed from the backplane) always comes back as 
    UNKNOWN CACHE TYPE. I moves between the two hsj controllers. At the
    suggestion of another engineer we attempted a 
    CLEAR_ERRORS INVALID_CACHE THIS_CONTROLLER NODESTROY_UNFLUSHED_DATA on
    the cache that showed as UNKNOWN CACHE TYPE. It now shows as read
    cache. I can take one controller who is set for 32MB write cache and
    leave him up., shutdown the other controller and reboot him. One that
    controller comes up it will have INVALID CACHE. The good controller
    will have READ CACHE. He have replaced one cache, replaced the new
    hsj40 spare without luck. 
    
      After much grief, he was able to get it to work by making both read
    cache, powering the hsj's off/on at the power supply and bring up into 
    dual-redundancy then shutting down and setting back to write back cache 
    and now all ok. I suspect we may have gotten into a loop when we made one 
    controller read cache and put it into a redundant configuration. 
    Then every time we did a failover copy we hosed the one controller.
    
      Does that soud plausible. Is UNKNOWN CACHE TYPE (a condition I had
    before making the one read cache) properly cleared with the command
    CLEAR_ERRORS INVALID_CACHE THIS_CONTROLLER NODESTROY_UNFLUSHED_DATA .
    
    
    Wes Gulden
    Colorado CSC
    CSC32::Gulden