Aruba & ProVision-based
1748146 Members
3569 Online
108758 Solutions
New Discussion

Linux bond auto failback not working correctly on two interconnected HP5400zl

 
TonyStocker
Occasional Contributor

Linux bond auto failback not working correctly on two interconnected HP5400zl

 

We have two 5400zl which we are using for a fault tolerant setup.

 

Switch A <-----ISL-----> Switch B

      ^                                             ^

       |                                              |

       |                                              |

       |                                              |

       ------- bond0 ----------

                           |

             +----------------+

              |    Server                |

            +----------------+

 

(hopefully my ascii art display correctly)

 

bond0 is made up of eth2 and eth7, mode is 1 (active-backukp), primary is eth2

 

eth2 is plugged into Switch-A Port a12, eth7 is plugged into Switch-B Port a12

 

The ISL is plugged into Switch-A Port a16 and Switch-B Port a16

 

When eth2 is unplugged, everything works as expected.  The bond fails over to eth7, and network traffic is basically unaffected (1 lost packet on a continuous ping.)

 

However when eth2 is plugged back into the switch, a problem arises.  On the server side, the bond fails back to eth2 since it's primary:

bonding: bond0: link status definitely down for interface eth2, disabling it
bonding: bond0: making interface eth7 the new active one.
bond0: link status definitely up for interface eth2, 1000 Mbps full duplex.
bonding: bond0: making interface eth2 the new active one.

 

However the continuous ping now hangs indefinitely UNTIL network activity on the server with the bond is initiated.  It's as if the network is unaware that traffic should now go back to Switch-A port a12.  I don't quite understand the problem, since I would have assumed that if there was any kind of routing issue here it would also show up on the initial failure of the link and the failover to eth7, but that works just fine. 

 

Is there some option or parameter on the switches that needs to be set?  Is there some way to determine why the traffic is "getting lost" on the switch?

1 REPLY 1
paulgear
Esteemed Contributor

Re: Linux bond auto failback not working correctly on two interconnected HP5400zl

Hi TonyStocker,

Did you have any luck in chasing down this issue? My guess would be that something's not quite the same in the setup of the two ports, but that's just a guess.

Does using Linux bond mode 5 (balance-tlb) or 6 (balance-alb) make any difference? Neither of those requires any special switch support, so they should work the same as active/standby with the added benefit of better performance.
Regards,
Paul