Putting compute nodes into maintenance mode
The maintenance mode is useful when you apply fixes or other upgrades to compute nodes.
Before you begin
You
must be assigned one of the following permissions to perform these
steps:
- System level, Hardware administration role with permission to Manage hardware resources (Full permission)
- Resource level administration role with Read access rights permission (ACL) to view a specific instance
- Resource level administration role with Write or All access rights permission (ACL) to view and manage a specific instance
About this task
Procedure
Results
Notes:
- When you evacuate a virtual machine instance from one compute node to another, the log file can include numerous DUPLICATE IP ADDRESS DETECTED messages. These messages are for informational purposes only and no action is required.
- Virtual machines that have been deployed within the last 3 hours cannot be evacuated to another compute node. Before you put a compute node into maintenance mode, wait at least 3 hours.
- When system-level high availability (HA) is enabled on the cloud group: if there is no space available for the virtual machines to be evacuated onto other compute nodes in the same cloud group and spare compute nodes are available, spare compute nodes are moved into the cloud group and the virtual machines are moved to the spare compute nodes.
- To put a compute node that is hosting a IBM Spectrum Scale server virtual machine into maintenance
mode (for example, so it can be removed from its cloud group), you
must first store the virtual machine. After stopping and storing the IBM Spectrum Scale server virtual machine and
quiescing the compute node, you can then restart the virtual machine
in a different compute node. When the stored IBM Spectrum Scale server virtual machine is started again, IBM Spectrum Scale starts automatically within
the virtual machine; there is no extra step to start IBM Spectrum Scale other than starting the server virtual
machine. Once the IBM Spectrum Scale server
virtual machine is no longer running on the compute node, you can
quiesce the compute node and put it in maintenance mode by following
the usual procedure.
Note that due to an anti-collocation constraint that IBM Spectrum Scale defines, all IBM Spectrum Scale virtual machines (except for the IBM Spectrum Scale Manager virtual machine) must be located on separate compute nodes. They also must be located in the same cloud group (this includes the IBM Spectrum Scale Manager virtual machine).