IBM SG24-5131-00 사용자 설명서

IBM Certification Study Guide  AIX HACMP

From time to time, it will be necessary to perform hardware maintenance or 
upgrades on cluster components. Some replacements or upgrades can be 
performed while the cluster is operative, while others require planned 
downtime. Make sure you plan all the necessary actions carefully. This will 
spare you a lot of trouble.

When maintaining or upgrading a node, cluster services must usually be 
stopped on the node. This means down time for the applications usually 
running on this node, at least during the takeover to other nodes. 

Consider the following points when replacing the whole or components of a 
node:

  • Make sure you have at least the same amount of RAM in the replacement 

system.

  • If your applications have been optimized for a particular processor or 

architecture, ensure that the new node is the same type of system. 
Uniprocessor applications may run slower on SMP systems.

  • Slot capacity of the new node must be the same or better.

  • Check the appropriate documentation for a proper adapter placement in 

your new node.

  • The license of your application may be dependent on the CPU ID. You 

may need to apply for a new license before trying to bring the new node 
into service.

In order to replace or add an adapter, the node must be powered off. This 
means down time for the applications usually running on this node, at least 
during the takeover to other nodes. 

Consider the following points when replacing or adding adapters in a node:

  • Make sure that the adapter is your problem and not faulty cabling. Bad 

cables are much more common than defective adapters. Most network 
and SSA cables can be changed online. Do some testing, for example, 
exchange the cables or try to connect to another port in your hub to see if 
the hub is your problem.