Resource-overflow-action behaviour with ADN



Resource-overflow-action behaviour is designed to bypass/drop new connections when the box is in memory pressure.

For an ADN deployment you would want this set up to bypass traffic rather than drop.


In the case of ADN you need to take account of whether explicit/translucent or transparent tunnels are being used. In the case of explicit/translucent tunnels

we preserve the tunnel so in the case of cifs or other long lived connections you may find that this is not a new tcp connection. 

When we talk about memory pressure we mean TM memory pressure and not the memory pressure configureable in health monitoring. You will know the box is in memory pressure if you see this eventlog message:

"TCP acceptance regulation: memory pressure is on"  0 30204:1   event_logger.cpp:31

The TM values which cannot be modified can be found at this link:



substitute x.x.x.x with the ip of your blue coat device.

By default they are:

Current_value:    31
 Upper limit:      80
 Lower limit:      75

It will be in memory pressure when the value goes over the upper limit and will come out of memory pressure when the value goes below the lower limit. You can see that the

current value is only 31 so not in memory pressure.

Now if the ADN box memory pressure goes above the "Upper limit" then it will notify its peers of its memory pressure. You will see this event log message on the peers

"5805.547 f=BCACHE p=EVENT m=TWM: Peer ProxySG 2607063345 entering mem-reg"  0 640002:1   debug_slog.hpp:325

Now if you try to run the "test adn" command from one of its peers you should see the following:

ProxySG VA 0000180276#test adn x.x.x.x 445
Connecting to x.x.x.x:445...succeeded!
Time            : 1.407 ms
Route decision  : Connect Direct
Route reason    : Bypassing ADN as all peers in memory regulation
Route policy    :
Connect result  : Failure
Failure reason  : Bypassing ADN as all peers in memory regulation
Network error   : No error(0)
ProxySG VA 0000180276#test adn x.x.x.x 445
Connecting to x.x.x.x:445...succeeded!

Now if clients behind the peer try to connect to the core which is in memory pressure the new connection will not be ADN optimized or compressed. They will still

receive protocol optimization (if the proxy can optimize that traffic) and benefit from object caching if the proxy provides it for that protocol. Now connections which

were set up before the memory pressure will still get ADN optimized until the session terminates. When the box is out of memory pressure the peers will be notified and all new connections will

receive ADN optimization (byte caching, compression).


Additional Information
Bug Number
InQuira Doc IdKB4866

Article Feedback

Did this Article solve your issue?
Additional Comments:
Previous MonthNext Month