|
Hi all,
I am currently facing a problem that I would like to discuss.
The Env :
I am testing VRRP implementation on a BIG ISP backbone (Thanks Paul for
your time). We try overload the network in order to test the VRRP framework
functionality on an overloaded env. Test is runing multiple "juno" client
simulating a big DOS and then we run multiple ping flood to charge more....
Then the network is very very loaded (linux router too).
The problem :
The network is congesting and some time (not regular frequency) some packet
are loose, if a VRRP MASTER advert is dropped, the remote BACKUPs said :
"Wow our master is down" => then a new master election is done.... But then
the remoteprevious master receive this advert and said : "Hey little boy,
this is me the BOSS" and previous master became again new MASTER.... Here
you can view the issue : The VRRP env is constently flapping from BACKUP
<->MASTER and this is nasty since some routings issues can be stalled and
performance decreased...
I was thinking of a "detecting network congestion" policy that change the
VRRP state machine if congestion is detected. In a way where BACKUP retry
receiving a new master advert on network congestion.
=> My question here is : What is the best way to detect network congestion
? (RTT, ...)
Best regards,
Alexandre
|
