Issue: Torque caches a lot of node information, including memory. After adding memory to a node, the server may not see that additional memory.
Solution: There is a fairly simple process, requiring restarting of pbs_server twice, which will cause Torque to update it's node configuration. These steps are:
- Shut down the pbs_server
- Remove the node(s) with additional memory from the "nodes" file (/var/spool/torque/server_priv/nodes, by default).
- "cd" to the /var/spool/torque/server_priv/node_usage directory and remove the file(s) whose name(s) correspond to the node(s) with new memory
- Start the pbs_server.
- Add the node(s) BACK into the "nodes" file.
- Restart the pbs_server.
After that second restart, Torque's view of the total memory for the node(s) should be correct.
Tags: memory, node_usage, upgrade.nodes