Why are NFSv4 callbacks not setup correctly in a High Available NFS-server environment?
Issue
In a clustered NFS environment, the NFSv4 callbacks are not setup correctly.
The NFS-client mounts an export on a fake NFS-server, the 'loadbalancer or director' in the diagram. This fake NFS-server forwards the connection on to one of the real NFS-servers.
+------------+
| NFS-client |
+------------+
| 192.168.130.24
|
|
| 192.168.130.23
+--------------+
| loadbalancer |
| or |
| director |
+--------------+
|
|
+--------------+-------------+
| |
| 192.168.130.20 | 192.168.130.21
+--------+--------+ +--------+--------+
| real NFS-server | | real NFS-server |
+-----------------+ +-----------------+
Now, when one of the read NFS-servers tries to setup the callback path to the NFS-client, the IP-address from the originating NFS-server is not the same as the one which the NFS-client used to mount the export. It seems that the NFS-client does not accept the CB_NULL
procedure call from the real NFS-servers, as the NFS-client does not send a reply on the procedure call.
Environment
- Red Hat Enterprise Linux 5, 6.0 and 6.1
-
mounting NFS exports over NFSv4 from a clustered NFS server
- the Hitachi NAS Platform 3090 is an example
Subscriber exclusive content
A Red Hat subscription provides unlimited access to our knowledgebase, tools, and much more.