• Ryan Jackson's avatar
    Handle failed XML-RPC calls · dc46b6ec
    Ryan Jackson authored
    If TestBusy() or ClearAddress() fails due to connection problems, the
    subboss_images table will be left in a bad state.  tmcd will assume
    that the frisbeed process is still running on the subboss and will
    return that address to nodes calling loadinfo.  These nodes will get
    stuck waiting for a frisbeed that isn't running.
    
    To prevent this, we now restart frisbeed when the TestBusy() or
    ClearAddress() calls fail.  Hopefully, the next time around the calls
    will succeed and subboss_frisbeelauncher can reset the DB state and
    exit normally.
    dc46b6ec
subboss_frisbeelauncher.in 11.2 KB