Major upgrade to the Compute (Nova) service.
The instance action
trigger_crash_dump has been added to the REST API and
will be available via the command line. This allows users to force
instances to perform a crash dump for debugging purposes by requesting a
non-maskable interrupt (NMI). Operating systems configured to respond to the
NMI will perform a core dump and reboot.
A custom description attribute can be added when creating, rebuilding or updating a server instance. In order to use this new feature, relevant request headers must specify at least microversion 2.19.
Events can now be retrieved for instances that have been deleted, which can be useful when performing an audit. It’s important to note that this only applies to instances deleted since this release.
This means that GET requests for server events, sent to the correct endpoint for the v2.1 API will still return an output even for instances that have been deleted as long as you are able to specify the UUID of said instance.
The EC2 compute API has now been configured to listen on port 8788 instead of port 8773. The updated API URLs can be found on the API page of the dashboard and in this section of the documentation.
Any Client software making use of the EC2 API for compute operations will need to be configured to use the new URL and port.
Please note note that this change does not affect EC2 credentials or object storage; these continue to operate as before.
Two new affinity policies are available this release; soft-affinity and soft-anti-affinity. These ‘soft’ policies are respectively similar to the existing affinity and anti-affinity policies, but differ from the existing policies in that they apply the (anti-)affinity rule as a preference rather than a strict rule. This difference is outlined in the table below:
Run all servers on the same hypervisor host. If this is not possible, do not start the servers that are unable to meet this policy.
Run all servers on separate hypervisor hosts. If this is not possible, do not start the servers that are unable to meet this policy.
Attempt to run all servers on the same hypervisor host. If this is not possible, start the remaining servers on as few other hypervisor hosts as possible.
In order to use this new feature, relevant request headers must specify at least microversion 2.15.
user-id are now included in the return
os-server-groups API requests. In order to use this new feature,
relevant request headers must specify at least microversion 2.13.
Some of the new features described above are only accessible from the v2.1
compute API. This is accessible from the same URL and port as the v2 API but
requires that the path is changed from
/v2.1/. So for example the
compute API v2.1 for the Porirua region is
For backwards compatibility reasons the service catalogue still publishes the v2 API, so in order to use the v2.1 API you will currently need to manually ove-ride this URL in the configuration of your cloud orchestration software.
Please note that the compute v2 API endpoint is being deprecated and will be retired after the next major compute upgrade.
The compute v2 API endpoint is being deprecated and will be retired after the next major compute upgrade.
The EC2 compute API on port 8773 is no longer available. This service now listens on port 8788.