CTDB failover with multiple VLAN tagged bonded interfaces don't work

[Robin Jonsson]

Hi Michael,
 
Thanks for getting back to me!
 
Here's the requested output:
 
Public IPs on node 0
192.168.4.4 node[0] active[bond0.123] available[bond0.123] configured[bond0.123]
192.168.4.72 node[-1] active[] available[] configured[bond0.124]
 
Public IPs on node 1
192.168.4.4 node[0] active[] available[bond0.123] configured[bond0.123]
192.168.4.72 node[-1] active[] available[] configured[bond0.124]
 
Interfaces on node 0
name:bond0.124 link:down references:0
name:bond0.123 link:up references:1
 
Interfaces on node 1
name:bond0.124 link:down references:0
name:bond0.123 link:up references:0
 
I raised the debug level to INFO and went through 'log.ctdb'. I still wasn't able to find anything obviously wrong. Only the first VIP gets assigned and it doesn't matter what order i specify the interfaces in public_addresses. Only the lowest one(bond0.123) gets assigned. If I remove bond0.123 and only use bond0.124, I can successfully online and use this VIP aswell...
 
2014/03/26 08:08:10.159072 [65054]: server/ctdb_takeover.c:162 public address '192.168.4.72' now unassigned (old iface '__none__' refs[0])
2014/03/26 08:08:10.159242 [65054]: server/ctdb_takeover.c:151 public address '192.168.4.4' now assigned to iface 'bond0.123' refs[1]
2014/03/26 08:08:10.159258 [65054]: Takeover of IP 192.168.4.4/26 on interface bond0.123
2014/03/26 08:08:10.159270 [65054]: server/eventscript.c:689 Starting eventscript takeip bond0.123 192.168.4.4 26
2014/03/26 08:08:10.290005 [65054]: server/eventscript.c:490 Eventscript takeip bond0.123 192.168.4.4 26 finished with state 0
2014/03/26 08:08:11.057499 [65104]: The interfaces status has changed on local node 0 - force takeover run
 
For a note: I have been able to reproduce this issue in a completely unrelated virtual environment aswell..

Thanks in advance! 

Regards,
Robin

On 26 Mar 2014, at 24:27 am, Michael Adam <obnox at samba.org> wrote:

> Hi Robin,
> 
> could you provide the output of "ctdb ip" and "ctdb ifaces"?
> Also, raising the debug level to INFO, might give a clue
> in the logs when starting CTDB.
> 
> Cheers - Michael
> 
> On 2014-03-25 at 19:00 +0000, Robin Jonsson wrote:
>> 
>> Hi, 
>> 
>> I have configured failover for my Gluster volumes using CTDB, which works just fine when I specify one or more IP address on the same VLAN tagged bonded interface like this: 
>> 
>> # cat /etc/ctdb/public_addresses
>> 192.168.4.4/26 bond0.123
>> 192.168.4.5/26 bond0.123
>> 
>> When I specify addresses from different VLANs (and subnets) like this, it doesn’t work:
>> 192.168.4.4/26 bond0.123
>> 192.168.4.72/26 bond0.124
>> 
>> … and I get ‘PARTIALLYONLINE’ when running ‘ctdb status’. The first specified interface goes online, the other one doesn’t.
>> 
>> # ctdb status
>> Number of nodes:2
>> pnn:0 192.168.4.2      PARTIALLYONLINE (THIS NODE)
>> pnn:1 192.168.4.3      PARTIALLYONLINE
>> Generation:624797803
>> Size:2
>> hash:0 lmaster:0
>> hash:1 lmaster:1
>> Recovery mode:NORMAL (0)
>> Recovery master:0
>> 
>> I have checked the logs, but I am unable to find anything that I’ve done obviously wrong.. 
>> 
>> Anyone got an idea?
>> 
>> Thanks in advance, 
>> Robin
> 
> 

-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 496 bytes
Desc: Message signed with OpenPGP using GPGMail
URL: <http://lists.samba.org/pipermail/samba-technical/attachments/20140326/44839c35/attachment.pgp>