Ok, my apologize, the stats in the PREROUTING chains seems to be
correct, they have 50% each:
root@VMWZLB003:/usr/local/zenloadbalancer/www# iptables -t nat -L -v -n
Chain PREROUTING (policy ACCEPT 1961 packets, 139K bytes)
pkts bytes target prot opt in out source
destination
835 49045 DNAT all -- * * 0.0.0.0/0
0.0.0.0/0 mark match 0x200 /* FARM_ZLB-EXCH-HUB_0_ */
to:192.168.120.1
828 48869 DNAT all -- * * 0.0.0.0/0
0.0.0.0/0 mark match 0x201 /* FARM_ZLB-EXCH-HUB_1_ */
to:192.168.120.2
If by chance you have a good document explaining this mangle probability
process, send it out, because i've been looking for explanation before
posting to ML but I couldnt find anything relevant.
Sorry for the noise then
tibz
On 13/12/2013 13:41, tibz wrote:
Hi,
Which one is the correct behavior?
Because in my setup both backend have a weight of 10, so loadbalancing
should be 50% each, correct?
If I look at the iptables output, I see that HUB_1 has probability of
1.000000 while HUB_0 has 0.500000
My understanding is that HUB_1 has 66% (1/1.5) of the load and HUB_0
has 33% (0.5/1.5) of the load. It seems to be confirmed by the stats
in the iptables output:
HUB_1 has received 335k packets for 417M, and HUB_0 has received 167k
packets for 208M.
So it seems that HUB_1 has received 2 times more request than HUB_0
Regards,
On 13/12/2013 13:13, Laura Garcia wrote:
Hi tibz, it's the correct behavior to calculate the probability based
in the desired weight.
Why do you suspect that is wrong?
Regards
El 13/12/2013 11:32, "tibz" <[email protected]
<mailto:[email protected]>> escribió:
Hello,
On zlb v3.03 i've configured a L4 DNAT farm. I've left the
loadbalancer
algorithm on "weight", and i've added 2 backend with each a
weight of 10.
That way I was expecting distribution to be 50% each.
root@VMWZLB003:/usr/local/zenloadbalancer/config# iptables -t
mangle -L
-v -n
Chain PREROUTING (policy ACCEPT 522K packets, 435M bytes)
pkts bytes target prot opt in out source destination
335K 417M MARK all -- * * 0.0.0.0/0
<http://0.0.0.0/0>
192.168.135.4 statistic mode random probability 1.000000 /*
FARM_ZLB-EXCH-HUB_1_ */ MARK set 0x201
167K 208M MARK all -- * * 0.0.0.0/0
<http://0.0.0.0/0>
192.168.135.4 statistic mode random probability 0.500000 /*
FARM_ZLB-EXCH-HUB_0_ */ MARK set 0x200
Obviously one is being given 2/3 (1 compared to 1.5) of
connections and
the other only get 1/3 (0.5 compared to 1.5).
If I set weight of 20 to the HUB_0, I have this result:
root@VMWZLB003:/usr/local/zenloadbalancer/config# iptables -t
mangle -L
-v -n
Chain PREROUTING (policy ACCEPT 510 packets, 177K bytes)
pkts bytes target prot opt in out source destination
117 115K MARK all -- * * 0.0.0.0/0
<http://0.0.0.0/0>
192.168.135.4 statistic mode random probability 1.000000 /*
FARM_ZLB-EXCH-HUB_1_ */ MARK set 0x201
83 87809 MARK all -- * * 0.0.0.0/0
<http://0.0.0.0/0>
192.168.135.4 statistic mode random probability 0.666667 /*
FARM_ZLB-EXCH-HUB_0_ */ MARK set 0x200
So HUB_O is not given twice the weight as it should...
Changing the algorith to Priority and setting a priority of 10 to
each
backend doesnt seems to be the solution. As far as I've seen,
priority
is only to have failover backend, not loadbalancing.
So do I misunderstand something or is the way you calculate the
weight
to set the probability in the iptables wrong?
I feel like it should be "probability = weight / sum(weight)", no?
Thanks
tibz
------------------------------------------------------------------------------
Rapidly troubleshoot problems before they affect your business.
Most IT
organizations don't have a clear picture of how application
performance
affects their revenue. With AppDynamics, you get 100% visibility
into your
Java,.NET, & PHP application. Start your 15-day FREE TRIAL of
AppDynamics Pro!
http://pubads.g.doubleclick.net/gampad/clk?id=84349831&iu=/4140/ostg.clktrk
_______________________________________________
Zenloadbalancer-support mailing list
[email protected]
<mailto:[email protected]>
https://lists.sourceforge.net/lists/listinfo/zenloadbalancer-support
------------------------------------------------------------------------------
Rapidly troubleshoot problems before they affect your business. Most IT
organizations don't have a clear picture of how application performance
affects their revenue. With AppDynamics, you get 100% visibility into your
Java,.NET, & PHP application. Start your 15-day FREE TRIAL of AppDynamics Pro!
http://pubads.g.doubleclick.net/gampad/clk?id=84349831&iu=/4140/ostg.clktrk
_______________________________________________
Zenloadbalancer-support mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/zenloadbalancer-support
------------------------------------------------------------------------------
Rapidly troubleshoot problems before they affect your business. Most IT
organizations don't have a clear picture of how application performance
affects their revenue. With AppDynamics, you get 100% visibility into your
Java,.NET, & PHP application. Start your 15-day FREE TRIAL of AppDynamics Pro!
http://pubads.g.doubleclick.net/gampad/clk?id=84349831&iu=/4140/ostg.clktrk
_______________________________________________
Zenloadbalancer-support mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/zenloadbalancer-support
------------------------------------------------------------------------------
Rapidly troubleshoot problems before they affect your business. Most IT
organizations don't have a clear picture of how application performance
affects their revenue. With AppDynamics, you get 100% visibility into your
Java,.NET, & PHP application. Start your 15-day FREE TRIAL of AppDynamics Pro!
http://pubads.g.doubleclick.net/gampad/clk?id=84349831&iu=/4140/ostg.clktrk
_______________________________________________
Zenloadbalancer-support mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/zenloadbalancer-support
