When the llttab of a VCS node is configured to use a file or llt hostid instead of the hostname, cfs commands such as 'cfscluster' can fail to recognise the correct hostname.
Article ID: 100022260
Updated On:
Description
Error Message
V-35-410: Cluster Server not running on local node: /etc/nodename
Resolution
As part of the SFCFS product, a suite of 'cfs' commands are provided to aid in administration. With 5.0MP3RP3 release threre were enhanced to enable cfs commands to work even when VCS is not running. As part of this enhancement a regression was introduced where these commands will fail if the 'set-node' line in the /etc/llttab file does not directly specify a hostname and instead either specifies a file or a llt hostid.
For example if the llttab file contains the line
set-node /etc/VRTSvcs/conf/sysname
Then commands such as
cfscluster status will fail with the following error: /opt/VRTS/bin/cfscluster status Error: V-35-410: Cluster Server not running on local node: /etc/VRTSvcs/conf/sysname A temporary workaround is to change the llttab file to use the nodename or modify the tr_lib.sh file to revert to the previous functionality of requiring VCS to be up when cfs commands are initiated.
To make this change edit file '/opt/VRTSvxfs/cfs/lib/tr_lib.sh' and change line 222 from
halocalnode=`cat /etc/llttab | grep "^set-node" | awk '{print $2}'`
to
halocalnode=`LANG=C hasys -nodeid 2>/dev/null | awk '{print $6}'`
A permanent fix has been created and is available by contacting Veritas support. The fix will be included in a future generally available patch.
Applies To
Issue is seen on 5.0MP3 and on 5.1SP1 versions.
Issue/Introduction
When the llttab of a VCS node is configured to use a file or llt hostid instead of the hostname, cfs commands such as 'cfscluster' can fail to recognise the correct hostname.
Was this article helpful?
Yes
No
Powered by
