Bug 7878 - Public ip isn't accessable after CTDB failover with bond device.
Public ip isn't accessable after CTDB failover with bond device.
Status: RESOLVED WORKSFORME
Product: CTDB 2.5.x or older
Classification: Unclassified
Component: ctdb
unspecified
x86 Linux
: P3 critical
: ---
Assigned To: Michael Adam
Samba QA Contact
:
Depends on:
Blocks:
  Show dependency treegraph
 
Reported: 2010-12-21 09:28 UTC by iron1982
Modified: 2016-09-02 00:52 UTC (History)
0 users

See Also:


Attachments

Note You need to log in before you can comment on or make changes to this bug.
Description iron1982 2010-12-21 09:28:11 UTC
hi, all.
I built my ctdb&samba cluster with 3 nodes. There is a bond interface which is consist of 4 network cards on each node. Then i configure 3 public ips for my cluster. 
After startup, each node got a public ip, and mount.cifs can access these public ips. My cifs client can read/write the shared file system.

Unfortunately, when i issue "ctdb disable" command on node1(whose's public ip originally is 192.168.1.204), node1 release its public ip, and node2 takes the ip. But my cifs client can't access 192.168.1.204 any more. Ping 92.168.1.204 failed.

I run arp on my cifs client, and found that node1's mac address matched 192.168.1.204. Then i found that my cifs client received arp packets from node1 every 2~3 seconds with tcpdump. 

192.168.1.204 is accessable again after i restart my cifs client's network.

can you tell me why node1 send arp to my cifs client after releasing the public ip? is that a bug?

BTW, i built another samba cluter with no bond device and it works well without that problem described above.

my english is poor. i hope you can understand me.

thank you.
Comment 1 Martin Schwenke 2016-09-02 00:52:05 UTC
We have seen SMB failover working correctly on bond interfaces for many years now.  I haven't seen this happen.  

I don't think we have enough information to understand what might have been happening on a system nearly 6 years ago, so I'm going to close this...