[Spread-users] spread slow to notice failures
Ryan Caudy
caudy at jhu.edu
Wed Apr 9 15:08:55 EDT 2003
To quote an earlier email sent by Yair:
"To me, the segment definition in your example below seems incorrect.
You should not include 127.0.0.1 in a spread configuration file unless
the configuration file includes exactly one segment with exactly one
machine."
Thus, the first thing I would try is following that advice.
--Ryan
Generic Player wrote:
> I am testing out spread right now to do load balancing between 2
> webservers, using wackamole. When I first start spread, they take
> about 10 seconds to find each other, and when I turn off one of the
> machines, it takes 30 seconds for the other to notice and respond.
>
> Wackamole is fine, as soon as I see the membership event from spread,
> wackamole moves the IPs, its just for some reason spread takes a long
> time to notice when the other machine goes down. I tried lowering the
> values in membership.c, but it had no effect. The only other thing
> I've seen is people saying can cause slowness like this is because of
> broadcast traffic not working, but broadcast traffic works fine from
> what I can tell.
>
> r.c and s.c don't compile, so I can't really test with those. But if I
> ping 10.0.0.255 both machines respond. The machines are directly
> connected to each other via crossover cable, 1000baseTX full-duplex.
> Network connectivity is fine, no errors, good throughput.
>
> Here's a copy of the config, its the same on both machines.
>
> Spread_Segment 127.0.0.255:4803 {
> localhost 127.0.0.1
> }
>
> Spread_Segment 10.0.0.255 {
> one 10.0.0.1
> two 10.0.0.2
> }
>
> _______________________________________________
> Spread-users mailing list
> Spread-users at lists.spread.org
> http://lists.spread.org/mailman/listinfo/spread-users
>
More information about the Spread-users
mailing list