Restarting services before a task is complete

Edit online

Starting, stopping, or restarting services when PowerVC tasks (such as deploying an image or attaching or detaching volumes) are running can cause unexpected results.

Explanation

When you perform tasks within PowerVC, those tasks might call on one or more OpenStack services when the tasks complete. For example, when you attach a volume, both the cinder and nova services are called in the successful completion of that task. If you stop or restart one of those OpenStack services while the volume is being attached, you might experience unexpected results, such as:
  • One of the resources might be put into a state that makes it unusable.
  • One or more resources that are involved might have incomplete tasks. This issue might result in miscommunication between the cinder and nova services that are related to that resource in the future.
Services can be stopped or restarted in the following ways:
  • Using the powervc-services command.
  • Performing certain tasks within PowerVC. The following table lists tasks that can cause services to stop or restart:
    Table 1. Actions that cause services to stop or restart.
    Action Service that gets stopped or restarted
    Edit storage (IP address, credentials, or display name) Cinder (volume)
    Edit the default storage template for a storage provider
    Edit storage switches (IP address, credentials, or display name)
    Remove a storage provider
    Edit HMC or Host (IP address, credentials, or display name) Nova (compute)
    Remove a host
    Running the powervc-ldap-config command Keystone (identity)
    Running the powervc-config command Varies, depending on the service being configured
    Performing a backup or restore All
    Note: Usually, PowerVC restricts you from performing an action when a task is not complete.

Resolution

Avoid restarting services while other PowerVC tasks are in progress.

However, if you run into problems because services are restarted, perform the following steps to resolve the resulting problems:
  • If the volumes or virtual machines are stuck in attaching or detaching state, you might have to recover the stuck volumes or virtual machines by using the recovery options. For more information about the recovery options, see Transient volume recovery.
  • If a task fails in the middle of the operation, attempt the task again.
  • Remove and re-add the affected resources from PowerVC.
  • For complex tasks, such as migration, you might have to roll back or undo the task outside of PowerVC and then perform the task again.