I have set up a bonding link with 2 wireless links, as shown below:
(PC1)–(RB750UP)==(2 x SXT)===wireless===(2 x SXT)==(RB750UP)–(PC2)
I have set up two EoIP tunnels to be able to bond over wireless, and it works nice; i can ping PC1 from PC2, do a btest, etc.
If I connect to one SXT and shutdown wlan interface, bonding is still up with half bandwidth, all is fine. So it seems to work as expected.
but
If i physically unplug one network cable from any SXT (simulating hardware failure), the whole bonding collapses.
I use balance-rr with ARP monitoring, am i doing something wrong or is it working as intended?
[test@BYA_SW03] > /interface bonding print
Flags: X - disabled, R - running
0 R name=“bonding-NB” mtu=1500 mac-address=FE:6B:20:14:64:AC arp=enabled slaves=eoip-5260,eoip-5300 mode=balance-rr primary=none link-monitoring=arp
arp-interval=100ms arp-ip-targets=192.168.10.190 mii-interval=100ms down-delay=0ms up-delay=0ms lacp-rate=30secs transmit-hash-policy=layer-2
I think if you are trying to use LAG (LACP) the Link aggregation group (LAG) will not collapses any more
this is really better in such case in my poor opinion.
Before, each “bonding member” was in a subnet with 4 IPs (Switch interface, SXT Emitter, SXT Station, Switch interface), with a EoIP tunnel connecting directly the two switches.
Now, i just removed all IPs and EoIP, bonding is working fine.
Got this idea from “Make sure that you do not have IP addresses on interfaces which will be enslaved for bonding interface!” in http://wiki.mikrotik.com/wiki/Manual:Interface/Bonding
So it looks like the problem was coming from the EoIP tunnel and/or the presence if IP addresses in the path (which is mandatory if you want to set up EoIP..). It’s working like this, i’m just a little worried because i can’t monitor my SXT devices as they don’t have any IP address anymore, well..
Thanks Kreacher for the tip, i have tried a little LACP ( = 802.3ad mode?), i didn’t get as good results as balance_rr mode, i will have a look again.
It is for bonding from 2 to a maximum of 8 ports (lines) and it is made exact for loosing a line (port)
and otherwise running with lower throughput but running!
I am not really fit in those things but as I am informed you could try out LAG (LACP) over round robin method
as well.