LAG problems

[wtm]

Seems we still have a problem with multiple LAGS between the Netonix switches and the Mikrotik routers. If we enable more than 2 LAG ports, it crashes the bonding connection and next to nothing goes through. This is using both 1.5.5 or 1.5.6 firmware. We have to roll back to 1.5.0 firmware for it to work. Mikrotik firmware is 6.46.4

I am going to try upgrading the Mikrotik firmware on the CCR unit tonight, and see if we have the same problem.

Another reason we need the 10 Gig switches !

[JustJoe]

Just wondering...

Do you have them configured as static or LACP ?

I'm looking at trying MikroTik routers and wondered if their true router interfaces support LACP (Cisco router interfaces do not AFAIK)

[wtm]

Have them setup as LACP-A

[wtm]

OK, additional information.........
Narrowed this down to between 1.5.0 and 1.5.3 firmware (1.5.0 works with multiple LAGS and 1.5.3 crashes when more than 2 LAGs are used)

It doesn't seem to matter what version firmware is in the Mikrotik router. We tried several (So it is NOT Mikrotik's problem)

I am assuming that what I read in the CHANGELOG is that the problem happened in 1.5.1 ?

Luckily we caught this issue now, as we ordered in 2 WS26 switches to use on our main hub as spares, and they can only use version 1.5.3 firmware or greater.
Which means we would IMMEDIATELY have had a problem with our network crashing.

So I plead with you !

Either get the 10 gig switches out or fix the LAG problem !

[wtm]

Narrowed it even further..........It crashes in the 1.5.1 version firmware !

[JustJoe]

I might be doing similar lab testing in the next few weeks...

Although it should not matter, are your LAGs aggregating

1) a VLAN trunk at each end
2) a single common VLAN at each end
3) non-VLAN ports
or
4) #2 at one end and # 3 at the other

[Ludvik]

I am not author of this thread. But I tried it on WS-12-250-DC 1.5.6 and RB 450Gx4 with 6.47.7

LACP-A on three ports, hash MAC/IP.
ROS bonding 802.3ad, hash policy layer 2 and 3

1) working (trunk on netonix, single vlan on ROS)
2) working
3) working

[wtm]

UPDATE:

It crashed with 802.3ad, but is working ok with "rr balanced" selected on the Mikrotik router. The Mikrotik's are CCR's (1009 and 1036) we tried various firmware on them, including the latest "long Term" and "Stable". The Netonix is a WS-26-500-DC with firmwares 1.5.0, 1.5.1,1.5.2,1.5.5, and 1.5.6 tried on them.

The 802.3ad selection seems to crash only after 3 or more LAG's are connected to the ports. (We are not using any VLAN's on them)

LACP-A on ports that were used. hash was MAC/IP (Basically default LAG setup on the Netonix unit, except for the LAG ID)
"802.3ad" mode crashes ONLY with firmware 1.5.1 or greater. I also tried LACP-P and STATIC, but neither one worked.

STP settings do not seem to make any differences.

Mikrotik setup was tried with "ALL" bonding modes, hash policy tried with "ALL" modes. "Layer 2 and 3" seem to give the most throughput though.
AGAIN, if LAG mode "rr balanced" is selected on the Mikrotik router it seems to work fine with more than 2 LAGs. (We are running 4 at this time)

We are running WS-12-DC units with 2 LAG's on them, and there does not seem to be a problem there. I haven't tried with additional LAG's on them yet.
We won't need more than 2 LAG's if we ever get the 10 gig switches to replace these with.