The order of the magic number in the VERITAS Volume Replicator (VVR) data packet header is reversed
Article ID: 100007782
Updated On:
Description
Error Message
Using tcpdump to analyze the network traffic show that the order of the magic number in the VVR data packet is reversed. The order is "4f 43 4d 4e" instead of the usual "4e 4d 43 4f", but the heartbeat packet order is normal.
Below is an example of a data packet:
0000 00 18 82 f9 b5 6a 00 18 82 f9 b5 59 08 00 45 00 .....j.....Y..E.
0010 05 dc 02 33 20 00 40 11 4d 7b 0a a9 f8 04 0a a9 ...3 .@.M{......
0020 f8 0c 9c f4 80 23 20 d8 b3 8a 13 7f 5b 6c 22 06 .....# .....[l".
0030 9b 66 4f 43 4d 4e 01 01 00 4d 50 e3 1c 00 a0 40 .fOCMN...MP....@
0040 00 00 a0 00 00 00 00 00 00 00 d0 20 00 00 00 00 ........... ....
0050 00 00 a7 10 82 23 03 00 00 00 00 00 00 00 00 00 .....#..........
Resolution
This is expected behavior. The magic number appears reversed on the wire (in the tcpdump output), but is rearranged appropriately before it reaches the nmcom layer. The reason we see the magic number reversed is due to Suse Linux machines use little-endian byte order, which can be verified as below:
# echo -n I | od -to2 | head -n1 | cut -f2 -d" " | cut -c6
1
Output 1 indicates a little-endian machine while output 0 indicates a big-endian machine.
The heartbeat packets do NOT report this checksum error in the messages logs "VxVM VVR vxio V-5-0-830 Header checksum". The reason for the magic number appearing as it is in the heartbeat packets is that they always use a 'packed' format - which is a host-independent format. The 'packed' format is not used for data packets if the primary and secondary hosts are on the same platform. Thus, the native format (little-endian) of Suse Linux is used for data packets, and we see the magic number reversed. This is handled by the OS and does not affect replication/nmcom layer in any way.
Applies To
SUSE 9.0
SF5.0MP4RP1
Issue/Introduction
When analyzing the network traffic, we can see that the order of the magic number in the VVR data packets header is reversed on Suse 9.0 with SF5.0MP4 installed. However this does not lead to a header checksum issue.
Was this article helpful?
Yes
No
Powered by
