LB server/proxy code does not check that client connection
is still live and no timeout fired during internal processing
of demanding queries (it continues after client disconnected
due to client-side timeout). This can represent considerable
unnecessary load on LB server machine.
This is becoming much more common and severe after
introduction of WM Replanner feature, I am reopening and increasing severity of this bug now.
This is a copy of original issue https://savannah.cern.ch/bugs/index.php?43705 by Zdenek Salvet
LB server/proxy code does not check that client connection is still live and no timeout fired during internal processing of demanding queries (it continues after client disconnected due to client-side timeout). This can represent considerable unnecessary load on LB server machine.
This is becoming much more common and severe after introduction of WM Replanner feature, I am reopening and increasing severity of this bug now.