diff --git a/api-guide/source/down_cells.rst b/api-guide/source/down_cells.rst new file mode 100644 index 0000000000..5d6bdf2959 --- /dev/null +++ b/api-guide/source/down_cells.rst @@ -0,0 +1,334 @@ +=================== +Handling Down Cells +=================== + +Starting from microversion 2.69 if there are transient conditions in a +deployment like partial infrastructure failures (for example a cell +not being reachable), some API responses may contain partial results +(i.e. be missing some keys). The server operations which exhibit this +behavior are described below: + +* List Servers (GET /servers): This operation may give partial + constructs from the non-responsive portion of the infrastructure. A + typical response, while listing servers from unreachable parts of + the infrastructure, would include only the following keys from + available information: + + - status: The state of the server which will be "UNKNOWN". + - id: The UUID of the server. + - links: Links to the servers in question. + + A sample response for a GET /servers request that includes one + result each from an unreachable and a healthy part of the + infrastructure is shown below. + + Response:: + + { + "servers": [ + { + "status": "UNKNOWN", + "id": "bcc6c6dd-3d0a-4633-9586-60878fd68edb", + "links": [ + { + "rel": "self", + "href": "http://openstack.example.com/v2/6f70656e737461636b20342065766572/servers/bcc6c6dd-3d0a-4633-9586-60878fd68edb" + }, + { + "rel": "bookmark", + "href": "http://openstack.example.com/6f70656e737461636b20342065766572/servers/bcc6c6dd-3d0a-4633-9586-60878fd68edb" + } + ] + }, + { + "id": "22c91117-08de-4894-9aa9-6ef382400985", + "name": "test_server", + "links": [ + { + "rel": "self", + "href": "http://openstack.example.com/v2/6f70656e737461636b20342065766572/servers/22c91117-08de-4894-9aa9-6ef382400985" + }, + { + "rel": "bookmark", + "href": "http://openstack.example.com/6f70656e737461636b20342065766572/servers/22c91117-08de-4894-9aa9-6ef382400985" + } + ] + } + ] + } + +* List Servers Detailed (GET /servers/detail): This operation may give + partial constructs from the non-responsive portion of the + infrastructure. A typical response, while listing servers from + unreachable parts of the infrastructure, would include only the + following keys from available information: + + - status: The state of the server which will be "UNKNOWN". + - id: The UUID of the server. + - tenant_id: The tenant_id to which the server belongs to. + - created: The time of server creation. + - links: Links to the servers in question. + + A sample response for a GET /servers/details request that includes + one result each from an unreachable and a healthy part of the + infrastructure is shown below. + + Response:: + + { + "servers": [ + { + "created": "2018-06-29T15:07:29Z", + "status": "UNKNOWN", + "tenant_id": "940f47b984034c7f8f9624ab28f5643c", + "id": "bcc6c6dd-3d0a-4633-9586-60878fd68edb", + "links": [ + { + "href": "http://openstack.example.com/v2.1/6f70656e737461636b20342065766572/servers/bcc6c6dd-3d0a-4633-9586-60878fd68edb", + "rel": "self" + }, + { + "href": "http://openstack.example.com/6f70656e737461636b20342065766572/servers/bcc6c6dd-3d0a-4633-9586-60878fd68edb", + "rel": "bookmark" + } + ] + }, + { + "OS-DCF:diskConfig": "AUTO", + "OS-EXT-AZ:availability_zone": "nova", + "OS-EXT-SRV-ATTR:host": "compute", + "OS-EXT-SRV-ATTR:hostname": "new-server-test", + "OS-EXT-SRV-ATTR:hypervisor_hostname": "fake-mini", + "OS-EXT-SRV-ATTR:instance_name": "instance-00000001", + "OS-EXT-SRV-ATTR:kernel_id": "", + "OS-EXT-SRV-ATTR:launch_index": 0, + "OS-EXT-SRV-ATTR:ramdisk_id": "", + "OS-EXT-SRV-ATTR:reservation_id": "r-y0w4v32k", + "OS-EXT-SRV-ATTR:root_device_name": "/dev/sda", + "OS-EXT-SRV-ATTR:user_data": "IyEvYmluL2Jhc2gKL2Jpbi9zdQplY2hvICJJIGFtIGluIHlvdSEiCg==", + "OS-EXT-STS:power_state": 1, + "OS-EXT-STS:task_state": null, + "OS-EXT-STS:vm_state": "active", + "OS-SRV-USG:launched_at": "2017-10-10T15:49:09.516729", + "OS-SRV-USG:terminated_at": null, + "accessIPv4": "1.2.3.4", + "accessIPv6": "80fe::", + "addresses": { + "private": [ + { + "OS-EXT-IPS-MAC:mac_addr": "aa:bb:cc:dd:ee:ff", + "OS-EXT-IPS:type": "fixed", + "addr": "192.168.0.3", + "version": 4 + } + ] + }, + "config_drive": "", + "created": "2017-10-10T15:49:08Z", + "description": null, + "flavor": { + "disk": 1, + "ephemeral": 0, + "extra_specs": { + "hw:cpu_policy": "dedicated", + "hw:mem_page_size": "2048" + }, + "original_name": "m1.tiny.specs", + "ram": 512, + "swap": 0, + "vcpus": 1 + }, + "hostId": "2091634baaccdc4c5a1d57069c833e402921df696b7f970791b12ec6", + "host_status": "UP", + "id": "569f39f9-7c76-42a1-9c2d-8394e2638a6d", + "image": { + "id": "70a599e0-31e7-49b7-b260-868f441e862b", + "links": [ + { + "href": "http://openstack.example.com/6f70656e737461636b20342065766572/images/70a599e0-31e7-49b7-b260-868f441e862b", + "rel": "bookmark" + } + ] + }, + "key_name": null, + "links": [ + { + "href": "http://openstack.example.com/v2.1/6f70656e737461636b20342065766572/servers/569f39f9-7c76-42a1-9c2d-8394e2638a6d", + "rel": "self" + }, + { + "href": "http://openstack.example.com/6f70656e737461636b20342065766572/servers/569f39f9-7c76-42a1-9c2d-8394e2638a6d", + "rel": "bookmark" + } + ], + "locked": false, + "metadata": { + "My Server Name": "Apache1" + }, + "name": "new-server-test", + "os-extended-volumes:volumes_attached": [], + "progress": 0, + "security_groups": [ + { + "name": "default" + } + ], + "status": "ACTIVE", + "tags": [], + "tenant_id": "6f70656e737461636b20342065766572", + "trusted_image_certificates": [ + "0b5d2c72-12cc-4ba6-a8d7-3ff5cc1d8cb8", + "674736e3-f25c-405c-8362-bbf991e0ce0a" + ], + "updated": "2017-10-10T15:49:09Z", + "user_id": "fake" + } + ] + } + + **Edge Cases** + + * **Filters:** If the user is listing servers using filters, results + from unreachable parts of the infrastructure cannot be tested for + matching those filters and thus no minimalistic construct will be + provided. Note that by default ``openstack server list`` uses the + ``deleted=False`` and ``project_id=tenant_id`` filters and since + we know both of these fundamental values at all times, they are + the only allowed filters to be applied to servers with only + partial information available. Hence only doing ``openstack + server list`` and ``openstack server list --all-projects`` (admin + only) will show minimalistic results when parts of the + infrastructure are unreachable. Other filters like ``openstack + server list --deleted`` or ``openstack server list --host xx`` + will skip the results depending on the administrator's + configuration of the deployment. Note that the filter ``openstack + server list --limit`` will also skip the results and if not + specified will return 1000 (or the configured default) records + from the available parts of the infrastructure. + + * **Marker:** If the user does ``openstack server list --marker`` it will + fail with a 500 if the marker is an instance that is no longer reachable. + + * **Sorting:** We exclude the unreachable parts of the infrastructure just like + we do for filters since there is no way of obtaining valid sorted results from + those parts with missing information. + + * **Paging:** We ignore the parts of the deployment which are non-responsive. + For example if we have three cells A (reachable state), B (unreachable state) + and C (reachable state) and if the marker is half way in A, we would get the + remaining half of the results from A, all the results from C and ignore cell B. + + .. note:: All the edge cases that are not supported for minimal constructs would + give responses based on the administrator's configuration of the deployment, + either skipping those results or returning an error. + +* Show Server Details (GET /servers/{server_id}): This operation may + give partial constructs from the non-responsive portion of the + infrastructure. A typical response while viewing a server from an + unreachable part of the infrastructure would include only the + following keys from available information: + + - status: The state of the server which will be "UNKNOWN". + - id: The UUID of the server. + - tenant_id: The tenant_id to which the server belongs to. + - created: The time of server creation. + - user_id: The user_id to which the server belongs to. This may be "UNKNOWN" + for older servers. + - image: The image details of the server. If it is not set like + in the boot-from-volume case, this value will be an empty string. + - flavor: The flavor details of the server. + - availability_zone: The availability_zone of the server if it was specified + during during boot time and "UNKNOWN" otherwise. + - power_state: Its value will be 0 (``NOSTATE``). + + A sample response for a GET /servers/{server_id} request that + includes one server from an unreachable part of the infrastructure + is shown below. + + Response:: + + { + "server": [ + { + "created": "2018-06-29T15:07:29Z", + "status": "UNKNOWN", + "tenant_id": "940f47b984034c7f8f9624ab28f5643c", + "id": "bcc6c6dd-3d0a-4633-9586-60878fd68edb", + "user_id": "940f47b984034c7f8f9624ab28f5643c", + "image": { + "id": "70a599e0-31e7-49b7-b260-868f441e862b", + }, + "flavor": { + "disk": 1, + "ephemeral": 0, + "extra_specs": { + "hw:cpu_policy": "dedicated", + "hw:mem_page_size": "2048" + }, + "original_name": "m1.tiny.specs", + "ram": 512, + "swap": 0, + "vcpus": 1 + }, + "OS-EXT-AZ:availability_zone": "geneva", + "OS-EXT-STS:power_state": 0, + } + ] + } + +* List Compute Services (GET /os-services): This operation may give + partial constructs for the services with :program:`nova-compute` as + their binary from the non-responsive portion of the + infrastructure. A typical response while listing the compute + services from unreachable parts of the infrastructure would include + only the following keys for the :program:`nova-compute` services + from available information while the other services like the + :program:`nova-conductor` service will be skipped from the result: + + - binary: The binary name of the service which would always be + ``nova-compute``. + - host: The name of the host running the service. + - status: The status of the service which will be "UNKNOWN". + + A sample response for a GET /servers request that includes two + compute services from unreachable parts of the infrastructure and + other services from a healthy one are shown below. + + Response:: + + { + "services": [ + { + "binary": "nova-compute", + "host": "host1", + "status": "UNKNOWN" + }, + { + "binary": "nova-compute", + "host": "host2", + "status": "UNKNOWN" + }, + { + "id": 1, + "binary": "nova-scheduler", + "disabled_reason": "test1", + "host": "host3", + "state": "up", + "status": "disabled", + "updated_at": "2012-10-29T13:42:02.000000", + "forced_down": false, + "zone": "internal" + }, + { + "id": 2, + "binary": "nova-compute", + "disabled_reason": "test2", + "host": "host4", + "state": "up", + "status": "disabled", + "updated_at": "2012-10-29T13:42:05.000000", + "forced_down": false, + "zone": "nova" + } + ] + } diff --git a/api-guide/source/index.rst b/api-guide/source/index.rst index cd6901bd69..972a030deb 100644 --- a/api-guide/source/index.rst +++ b/api-guide/source/index.rst @@ -85,6 +85,7 @@ Contents paginated_collections polling_changes request_and_response_formats + down_cells .. toctree:: :hidden: diff --git a/api-guide/source/server_concepts.rst b/api-guide/source/server_concepts.rst index aa6c600552..a4a7acb6fd 100644 --- a/api-guide/source/server_concepts.rst +++ b/api-guide/source/server_concepts.rst @@ -77,8 +77,9 @@ server status is one of the following values: Suspending a server is similar to placing a device in hibernation; memory and vCPUs become available to create other servers. -- ``UNKNOWN``: The state of the server is unknown. Contact your cloud - provider. +- ``UNKNOWN``: The state of the server is unknown. It could be because a part + of the infrastructure is temporarily down (see :doc:`down_cells` + for more information). Contact your cloud provider. - ``VERIFY_RESIZE``: System is awaiting confirmation that the server is operational after a move or resize. diff --git a/api-ref/source/os-services.inc b/api-ref/source/os-services.inc index 477e016bc0..55d1216f3d 100644 --- a/api-ref/source/os-services.inc +++ b/api-ref/source/os-services.inc @@ -21,6 +21,14 @@ Lists all running Compute services. Provides details why any services were disabled. +.. note:: Starting with microversion 2.69 if service details cannot be loaded + due to a transient condition in the deployment like infrastructure failure, + the response body for those unavailable compute services in the down cells + will be missing keys. See `handling down cells + `__ + section of the Compute API guide for more information on the keys that + would be returned in the partial constructs. + Normal response codes: 200 Error response codes: unauthorized(401), forbidden(403) diff --git a/api-ref/source/servers.inc b/api-ref/source/servers.inc index 42beb1d1bc..6e2bd07ff3 100644 --- a/api-ref/source/servers.inc +++ b/api-ref/source/servers.inc @@ -175,6 +175,14 @@ whitelist will be silently ignored. - For admin user, whitelist includes all filter keys mentioned in :ref:`list-server-request` Section. +.. note:: Starting with microversion 2.69 if server details cannot be loaded + due to a transient condition in the deployment like infrastructure failure, + the response body for those unavailable servers will be missing keys. See + `handling down cells + `__ + section of the Compute API guide for more information on the keys that + would be returned in the partial constructs. + Normal response codes: 200 Error response codes: badRequest(400), unauthorized(401), @@ -506,6 +514,14 @@ OS-SRV-USG:terminated_at attributes. HostId is unique per account and is not globally unique. +.. note:: Starting with microversion 2.69 if server details cannot be loaded + due to a transient condition in the deployment like infrastructure failure, + the response body for those unavailable servers will be missing keys. See + `handling down cells + `__ + section of the Compute API guide for more information on the keys that + would be returned in the partial constructs. + Normal response codes: 200 Error response codes: badRequest(400), unauthorized(401), @@ -665,6 +681,14 @@ HostId is unique per account and is not globally unique. The server must exist. +.. note:: Starting with microversion 2.69 if the server detail cannot be loaded + due to a transient condition in the deployment like infrastructure failure, + the response body for the unavailable server will be missing keys. See + `handling down cells + `__ + section of the Compute API guide for more information on the keys that + would be returned in the partial constructs. + Normal response codes: 200 Error response codes: unauthorized(401), forbidden(403), diff --git a/doc/source/user/cells.rst b/doc/source/user/cells.rst index 7e79d47fff..d0fc7d5892 100644 --- a/doc/source/user/cells.rst +++ b/doc/source/user/cells.rst @@ -806,3 +806,17 @@ FAQs solved by adding a generic by-aggregate filter, which would allow listing instances on hosts contained within any aggregate, including one that matches the cell boundaries if so desired. + +#. Why are the API responses for ``GET /servers``, ``GET /servers/detail``, + ``GET /servers/{server_id}`` and ``GET /os-services`` missing some + information for certain cells at certain times? Why do I see the status as + "UNKNOWN" for the servers in those cells at those times when I run + ``openstack server list`` or ``openstack server show``? + + Starting from microversion 2.69 the API responses of ``GET /servers``, + ``GET /servers/detail``, ``GET /servers/{server_id}`` and + ``GET /os-services`` may contain missing keys during down cell situations. + See the `Handling Down Cells + `__ + section of the Compute API guide for more information on the partial + constructs. \ No newline at end of file diff --git a/nova/api/openstack/compute/rest_api_version_history.rst b/nova/api/openstack/compute/rest_api_version_history.rst index dadd19b567..a1a331b6df 100644 --- a/nova/api/openstack/compute/rest_api_version_history.rst +++ b/nova/api/openstack/compute/rest_api_version_history.rst @@ -865,3 +865,17 @@ be used to specify cinder ``volume_type`` when creating a server. ---- Remove support for forced live migration and evacuate server actions. + +2.69 +---- + +Add support for returning minimal constructs for ``GET /servers``, +``GET /servers/detail``, ``GET /servers/{server_id}`` and ``GET /os-services`` +when there is a transient unavailability condition in the deployment like an +infrastructure failure. Starting from this microversion, the responses from the +down part of the infrastructure for the above four requests will have missing +key values to make it more resilient. The response body will only have a +minimal set of information obtained from the available information in the API +database for the down cells. See `handling down cells +`__ for +more information. diff --git a/nova/conf/api.py b/nova/conf/api.py index 3be1819120..9395330612 100644 --- a/nova/conf/api.py +++ b/nova/conf/api.py @@ -340,7 +340,9 @@ True. Note that from API microversion 2.69 there could be transient conditions in the deployment where certain records are not available and the results could be partial for certain requests containing those records. In those cases this -option will be ignored. +option will be ignored. See "Handling Down Cells" section of the Compute API +guide (https://developer.openstack.org/api-guide/compute/down_cells.html) for +more information. """), ] diff --git a/releasenotes/notes/bp-handling-down-cell-10f76145d767300c.yaml b/releasenotes/notes/bp-handling-down-cell-10f76145d767300c.yaml new file mode 100644 index 0000000000..81a78833c1 --- /dev/null +++ b/releasenotes/notes/bp-handling-down-cell-10f76145d767300c.yaml @@ -0,0 +1,12 @@ +--- +features: + - | + From microversion 2.69 the responses of ``GET /servers``, + ``GET /servers/detail``, ``GET /servers/{server_id}`` and + ``GET /os-services`` will contain missing keys during down cell situations + because of adding support for returning minimal constructs based on the + available information from the API database for those records in the + down cells. See `Handling Down Cells`_ for more information on the missing + keys. + + .. _Handling Down Cells: https://developer.openstack.org/api-guide/compute/down_cells.html \ No newline at end of file