Views:

Summary



When running an NDMP backup using the DPX nibbler client interface from a DPX NDMP client to a DPX NDMP device server, the backup fails during drive allocation.

Symptoms



The following messages are reported in the job log:

SNBTMM5461O Node(MyDeviceServer) drive(LTO_0_007) volser(G01011) partition(14) mount rc=(0)

SNBJH_3036E *** Drive fail reported: drive LTO_0_007 on node MyDeviceServer in job 1107231314 ***

SNBEHB3021E 3372: *** Fail searching socket 5 of module ssdmbs at node [dev ] ***

On the client side, the following error messages are in the nibbler log (nibbler.log) in the BEX logs folder:

MyClient BackupExpressNibbler[15823]: DataConnect: Failed connecting to:[:10.1.1.1] port:[29459]. System call connect(6) failed: [strerror(146):Connection refused].:

In this case, the device server knows itself by name through the 10.1.1.1 interface. However, the client cannot address the device server on the 10.1.1.1 interface, though it knows the device server through a different IP address, e.g. 192.1.1.1, and a route is not known through the 10.1.1.1 interface Both machines can ping each other by name, though the respective interface used to do so on each machine is different.

 



Resolution



The DPX nibbler module performs the communication that is relevant to this problem. In this case, the device server nibbler is passing the IP address that it is listening on for the incoming data connection to the client side nibbler module. Note that the client side nibbler module will listen on all interfaces, so this message can be received, but the data connection will be attempted on that same interface, which is not desirable in this case. When the data connection is attempted, the passed IP address from the device server is unknown and the message connection refused is the result.

An update is available that enables a listing of IP addresses sent by the nibbler module, which are the interfaces that this server can be addressed by. This IP address list is available only in the NDMP version 4 protocol. When the client receives this IP address list, it will attempt to make connections to the server on each IP address in the list. The first address in the list is attempted first, and all addresses will be checked until a connection is successfully made. On the first successful connection, the data connection is established on that interface.

To enable this change, environment settings as well as an updated nibbler module are required on machines using the Advanced Protection manager features. First, you must force the DPX nibbler interface to run the NDMP version 4 protocol. Then, you must specify the listing of IP addresses that each server is known as. This is accomplished by:

Windows:

On all machines using the Advanced Protection manager features, create the following string value:

HKLM\Software\Syncsort\BackupExpress\<Machine Name>\0\NIB_MAXNDMPVER

With the following value data of the NDMP protocol to use:

4

Then, create the following string value:

HKLM\Software\Syncsort\BackupExpress\<Machine Name>\0\NIB_LISTEN_IP_LIST

With the following value data:

ADDR1:ADDR2:::ADDRN

where the IP address list is the value data here separated by colons. Finally, stop and start the DPX Advanced Protection Manager service.

 

UNIX:

Modify the startup script bexads in the misc directory of DPX. On machines using the Advanced Protection Manager features, export the following two environment variables at the beginning of the script:

exportNIB_MAXNDMPVER=4

In addition, export the IP address list separated by colons.

exportNIB_LISTEN_IP_LIST=ADDR1:ADDR2:::ADDRN

Then, stop and start nibbler, via bexads stop and bexads start. As an example, the IP address list for the device server mentioned in the symptom of this article should read as follows: 10.1.1.1:192.1.1.1

Support for this change has been addressed in an "All Servers" update. The updates can be found on the mysupport.catalogicsoftware.com. Contact Technical Support if you do not find the update for your platform.