Mailing List Archive

heartbeat protocol progress
Hi,

I added code to allow heartbeat to simulate packet loss to whatever degree
desired. This new code is now in CVS.

I've been running for about 15 hours continuously with a 25% packet loss rate on
transmit and receive for one of the two machines.

Strangely enough, it seems to have been behaving exactly as I hoped it would.

Each machine has seen about 18000 lost data packets, and they were retransmitted
and received correctly.

At this point, I'd say the code looks like it has basic sanity.

I suppose the next step is to try and raise the loss percentage to something
like 90% and see how it behaves.

There is a threshold in the code where it can fail to be able to retransmit a
packet. I suspect that 90% packet loss on transmit and receive will likely
reach this pretty quickly... I'll keep you up to date.

Instructions for testers:

Create a new file called /etc/ha.d/OnlyForTesting. Put these contents in it:
rcvloss=0.25
xmitloss=0.25

The syntax is very simple, but not forgiving. Don't put any blanks, and spell
everything exactly as stated...


Thanks!

-- Alan Robertson
alanr@bell-labs.com