-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 Erkki Lintunen wrote: | Hi, | | I'll do my best to describe the problem. I tested the config manually by | running following commands: | | # insmod bonding mode=1 arp_interval=100 | arp_ip_targets=10.0.3.1,10.0.3.2 primary=eth1 | # insmod 8021q | # insmod e100 | # insmod e1000 | | The module load order doesn't affect the problem here. It is now heavily | tested and the above is the recommended one in bonding documentation.
I've not personally used bonding in mode 1, I use the 802.3ad link aggregation mode, which seems to be more 'automatic'. Is there some reason you're not using this protocol? I believe the Procurve switches you have support it. <snip> | When I add an vlan onto the bond interface: | | # vconfig set_name_type VLAN_PLUS_VID_NO_PAD | # vconfig add bond0 3 | | "3" is the VLAN tag ID, ie. from 1 to 4096, in practice some switches | and/or switch vendors restrict available ID space ie. either you can | have say max 8 vlans defined in one switch but the switch should not | retrict 4096 IDs flowing through it or some vendors reserve IDs for some | special purposes. This way "600" is valid VLAN ID. | | Now disabling the primary interface's (ie. eth1) peer on the switch also | halts ping from 10.0.3.1 to 10.0.3.3, BUT when | | # ip addr add 10.0.4.1/24 dev vlan3 | # ip link set vlan3 up | | ping from 10.0.4.2 to 10.0.4.1 flows smoothly. Enabling the peer | interface of eth1 restores ping from 10.0.3.1 to 10.0.3.3 again. | | In other words the active backup interface (eth0 in this setup) becomes | active and the bond interface has vlans, the vlans continue to work but | the native non-vlan traffic blocks. The similar setup on latest kernels | with bonding 3.5.x work fine with the same switches. Hmm...other than upgrading your kernel, which would apparently fix your problem, have you checked your arp settings on both ends? If the 10.0.3.3/24 IP on bond0 is somehow getting the MAC address of the primary interface, and not the virtual MAC address of bond0, then taking down eth1 would cause ping to stop working (at least until the arp-cache entry timed out). I also assume you have your switch properly setup to bond the two interfaces? Finally, is there any difference in the contents of /proc/net/bonding/bond0 when running with a kernel that works vs the leaf kernel? - -- Charles Steinkuehler [EMAIL PROTECTED] -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.4.7 (MingW32) Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org iD8DBQFIMxn4LywbqEHdNFwRAtkaAKCVXHkt/tM8pH1PMqAQExROmiQzuwCfSkjs 3hg81EB2yALmNh84zXbe/bA= =Vem3 -----END PGP SIGNATURE----- ------------------------------------------------------------------------- This SF.net email is sponsored by: Microsoft Defy all challenges. Microsoft(R) Visual Studio 2008. http://clk.atdmt.com/MRT/go/vse0120000070mrt/direct/01/ ------------------------------------------------------------------------ leaf-user mailing list: leaf-user@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/leaf-user Support Request -- http://leaf-project.org/
