)]}'
{"/PATCHSET_LEVEL":[{"author":{"_account_id":4690,"name":"melanie witt","display_name":"melwitt","email":"melwittt@gmail.com","username":"melwitt"},"change_message_id":"9b57b7d11ae30f8119e26d4d59e36ec537afcf88","unresolved":false,"context_lines":[],"source_content_type":"","patch_set":1,"id":"0c3a571e_558b9dbd","updated":"2026-11-17 17:01:41.000000000","message":"We discussed in the IRC meeting today and agreed this is not the approach we want.\n\nRather than depart from the standard RPC queue naming convention, we should rather do something more like Nova where both hostnames are represented for the device (i.e. host and hypervisor_hostname).\n\nAs this issue was raised by CI running on the nova-operator repo [1], something to note is that for the Nova service, the deployment sets the `CONF.host` specifically to the FQDN [2] which is calculated by the openstack-operator [3].\n\nMaybe the right thing to do is just do the same thing for Cyborg?\n\n[1] https://github.com/openstack-k8s-operators/nova-operator\n[2] https://github.com/openstack-k8s-operators/edpm-ansible/blob/0fa6f9bfd7996deb5eb2b4c0efdf54d525bdc460/roles/edpm_nova/templates/config/02-nova-host-specific.conf.j2#L3\n[3] https://github.com/openstack-k8s-operators/openstack-operator/blob/be32086a80997ba883f3832c5728e061bfdeafc0/internal/dataplane/inventory.go#L293-L301","commit_id":"9360d375a276eaac7e7ca2c61b0d8e9da9c41f35"},{"author":{"_account_id":4690,"name":"melanie witt","display_name":"melwitt","email":"melwittt@gmail.com","username":"melwitt"},"change_message_id":"0c9976be7d8f807ebc34c4abb19048c604241eb7","unresolved":true,"context_lines":[],"source_content_type":"","patch_set":1,"id":"18ada219_4a92a3dd","in_reply_to":"0c3a571e_558b9dbd","updated":"2026-11-17 17:02:19.000000000","message":"(didn\u0027t mean to post this as Resolved)","commit_id":"9360d375a276eaac7e7ca2c61b0d8e9da9c41f35"},{"author":{"_account_id":4690,"name":"melanie witt","display_name":"melwitt","email":"melwittt@gmail.com","username":"melwitt"},"change_message_id":"6c7735054ee5b15c85bfe0de2be7e4eb79544348","unresolved":true,"context_lines":[],"source_content_type":"","patch_set":1,"id":"2dc6d366_754272a1","in_reply_to":"18ada219_4a92a3dd","updated":"2026-11-17 17:49:52.000000000","message":"\u003e Maybe the right thing to do is just do the same thing for Cyborg?\n\nUpdate: while it is true that Nova sets the `CONF.host` in that manner, it is not required for Nova RPC to work properly because Nova always uses the `CONF.host` for RPC targeting whereas in Cyborg it is not. Cyborg RPC is tied to the resource provider name currently.\n\nSo the only real fix seems to be to store both names like Nova does and use the FQDN only when it\u0027s needed.","commit_id":"9360d375a276eaac7e7ca2c61b0d8e9da9c41f35"}],"cyborg/api/controllers/v2/deployables.py":[{"author":{"_account_id":28006,"name":"teim-ci","display_name":"teim-ci","email":"ci@seanmooney.info","username":"ci-sean-mooney","status":"this is a third-party ci account run by sean-k-mooney on irc\nhosted at zuul.teim.app"},"tag":"autogenerated:zuul:automatic-ci","change_message_id":"76d3e7cdffa2e0aeb61b0862dc33c5c5bfea4afe","unresolved":false,"context_lines":[{"line_number":155,"context_line":"        # carries one. The device hostname is the nova compute"},{"line_number":156,"context_line":"        # resource provider name, which is not necessarily the name"},{"line_number":157,"context_line":"        # the agent serves RPC under."},{"line_number":158,"context_line":"        service_host \u003d ("},{"line_number":159,"context_line":"            obj_dev.service_host"},{"line_number":160,"context_line":"            if obj_dev.obj_attr_is_set(\u0027service_host\u0027)"},{"line_number":161,"context_line":"            else None"}],"source_content_type":"text/x-python","patch_set":2,"id":"fc72bfa7_02a35bed","line":158,"updated":"2026-11-17 19:00:46.000000000","message":"The program controller now resolves the agent target from device.service_host with a hostname fallback, but the only controller-level tests (cyborg/tests/unit/api/controllers/v2/test_fpga_program.py) use a fake Device without service_host and never assert the hostname argument passed to AgentAPI.fpga_program; the file contains no assert_called* at all. The conductor-side equivalent in FPGAExtARQ._do_programming gained new tests, so a future regression confined to this controller\u0027s resolution would pass CI while reintroducing the MessagingTimeout this change fixes.\n\n**Severity**: WARNING | **Confidence**: 0.85\n\n**Impact**: A regression in the controller\u0027s service_host resolution (wrong precedence, dropped fallback, renamed field) would silently reintroduce the fixed failure -- fpga_program published to a routing key with no bound queue, blocking the program API until rpc_response_timeout -- with no test failure to catch it.\n\n**Suggestion**:\nExtend test_fpga_program.py: in one test return a fake device carrying service_host and assert mock fpga_program was called with that host; keep another test with service_host absent asserting the hostname fallback is passed.","commit_id":"f734d68ef218443e780e0b652a6e181e6aee60f5"},{"author":{"_account_id":4690,"name":"melanie witt","display_name":"melwitt","email":"melwittt@gmail.com","username":"melwitt"},"change_message_id":"a559333feec81b3a17165999cf86e3a893cd1b3f","unresolved":false,"context_lines":[{"line_number":155,"context_line":"        # carries one. The device hostname is the nova compute"},{"line_number":156,"context_line":"        # resource provider name, which is not necessarily the name"},{"line_number":157,"context_line":"        # the agent serves RPC under."},{"line_number":158,"context_line":"        service_host \u003d ("},{"line_number":159,"context_line":"            obj_dev.service_host"},{"line_number":160,"context_line":"            if obj_dev.obj_attr_is_set(\u0027service_host\u0027)"},{"line_number":161,"context_line":"            else None"}],"source_content_type":"text/x-python","patch_set":2,"id":"be140276_9b01db3e","line":158,"in_reply_to":"fc72bfa7_02a35bed","updated":"2026-11-17 19:14:22.000000000","message":"Done","commit_id":"f734d68ef218443e780e0b652a6e181e6aee60f5"}],"cyborg/common/service.py":[{"author":{"_account_id":28006,"name":"teim-ci","display_name":"teim-ci","email":"ci@seanmooney.info","username":"ci-sean-mooney","status":"this is a third-party ci account run by sean-k-mooney on irc\nhosted at zuul.teim.app"},"tag":"autogenerated:zuul:automatic-ci","change_message_id":"d0d7b6eb5dce0080862d694c10ff90288711fc15","unresolved":false,"context_lines":[{"line_number":64,"context_line":"            periodic_interval_max\u003dCONF.periodic_interval,"},{"line_number":65,"context_line":"        )"},{"line_number":66,"context_line":""},{"line_number":67,"context_line":"        LOG.info("},{"line_number":68,"context_line":"            \u0027Created RPC server for service %(service)s on host %(host)s.\u0027,"},{"line_number":69,"context_line":"            {\u0027service\u0027: self.topic, \u0027host\u0027: server},"},{"line_number":70,"context_line":"        )"}],"source_content_type":"text/x-python","patch_set":1,"id":"324dbcbd_53ad0d31","line":67,"updated":"2026-11-17 01:02:47.000000000","message":"The change updates the \u0027Created RPC server\u0027 message to log the resolved server name, but the matching \u0027Stopped RPC server\u0027 message in RPCService.stop() still logs self.host. On agent deployments where the validated resource provider name differs from CONF.host - the deployments this patch targets - startup logs \u0027on host \u003crp-name\u003e\u0027 while shutdown logs \u0027on host \u003cconf-host\u003e\u0027, so the paired entries disagree for the same service instance.\n\n**Severity**: SUGGESTION | **Confidence**: 0.85\n\n**Impact**: Operators correlating agent start/stop entries, or searching logs for the queue/host name the agent actually serves, will not find the matching shutdown entry on deployments where the names differ, complicating restart and incident analysis.\n\n**Recommendation**:\nStore the resolved name on the service (for example `self.server \u003d server` in start()) and use it in the stop() log message so both messages report the name the RPC server actually bound under.","commit_id":"9360d375a276eaac7e7ca2c61b0d8e9da9c41f35"}],"cyborg/objects/device.py":[{"author":{"_account_id":28006,"name":"teim-ci","display_name":"teim-ci","email":"ci@seanmooney.info","username":"ci-sean-mooney","status":"this is a third-party ci account run by sean-k-mooney on irc\nhosted at zuul.teim.app"},"tag":"autogenerated:zuul:automatic-ci","change_message_id":"12ffb27f9a20652c3948fd90aacde5e6c0fdb9b3","unresolved":false,"context_lines":[{"line_number":34,"context_line":"    # Version 1.3: Add MDEV, PCI type"},{"line_number":35,"context_line":"    # Version 1.4: Add NVME type"},{"line_number":36,"context_line":"    # Version 1.5: Add device_state field"},{"line_number":37,"context_line":"    # Version 1.6: Add service_host field"},{"line_number":38,"context_line":"    VERSION \u003d \u00271.6\u0027"},{"line_number":39,"context_line":""},{"line_number":40,"context_line":"    dbapi \u003d dbapi.get_instance()"}],"source_content_type":"text/x-python","patch_set":3,"id":"cdd09190_11704b3c","line":37,"updated":"2026-11-17 19:30:24.000000000","message":"The change bumps Device VERSION to 1.6 and adds the service_host field but does not extend Device.obj_make_compatible to pop service_host when serializing for a target version below 1.6. Every prior field addition in this object follows the strip pattern (device_state below 1.5, status below 1.2), and DriverDevice handles this very field with \u0027if target_version \u003c (1, 5): primitive.pop(\u0027service_host\u0027, None)\u0027. The downgrade contract for the new field is therefore enforced inconsistently between the two objects this patch changes.\n\n**Severity**: SUGGESTION | **Confidence**: 0.85\n\n**Impact**: Any future caller that serializes a Device with target_version \u003c 1.6 (obj_to_primitive(target_version\u003d...)) leaks the new field to consumers registered at older object versions, and the downgrade behavior of service_host silently diverges between Device and DriverDevice, inviting copy-paste drift in the next field addition.\n\n**Recommendation**:\nAdd \u0027if target_version \u003c (1, 6): primitive.pop(\u0027service_host\u0027, None)\u0027 to Device.obj_make_compatible and extend cyborg/tests/unit/objects/test_device.py with a drop-below-1.6 assertion matching the existing device_state and status tests.","commit_id":"6050d3c4534568aba9372b89f78e19c70472ba7f"}],"cyborg/objects/extarq/fpga_ext_arq.py":[{"author":{"_account_id":28006,"name":"teim-ci","display_name":"teim-ci","email":"ci@seanmooney.info","username":"ci-sean-mooney","status":"this is a third-party ci account run by sean-k-mooney on irc\nhosted at zuul.teim.app"},"tag":"autogenerated:zuul:automatic-ci","change_message_id":"1ba3fedd69c62671440e2088c4c0076b68bb8e9a","unresolved":false,"context_lines":[{"line_number":249,"context_line":"            if device.obj_attr_is_set(\u0027service_host\u0027)"},{"line_number":250,"context_line":"            else None"},{"line_number":251,"context_line":"        )"},{"line_number":252,"context_line":"        hostname \u003d service_host or self.arq.hostname"},{"line_number":253,"context_line":"        driver_name \u003d deployable.driver_name"},{"line_number":254,"context_line":""},{"line_number":255,"context_line":"        # query_filter \u003d {\"device_id\": deployable.device_id}"}],"source_content_type":"text/x-python","patch_set":6,"id":"2b019627_edddcf0f","line":252,"updated":"2026-11-18 19:57:28.000000000","message":"The change establishes device.service_host as the agent RPC identity and applies it to two server-addressed agent RPC callers (DeployablesController.program and FPGAExtARQ._do_programming). The other two callers of the same RPC class, ExtARQ._allocate_attach_handle (create_vgpu_mdev) and ExtARQ.unbind via _deallocate_attach_handle (remove_vgpu_mdev), still derive their target from arq.hostname. Per the commit message, arq.hostname is the nova compute node (resource provider) name, which on the affected deployments is not the name the agent serves RPC under. The identical MessagingTimeout failure this change closes remains on the MDEV bind/unbind path.\n\n**Severity**: WARNING | **Confidence**: 0.85\n\n**Impact**: On deployments where CONF.host differs from the resource provider name (the deployments this change targets), binding a vGPU accelerator still publishes create_vgpu_mdev to a routing key with no bound queue and blocks until rpc_response_timeout, so the bind job fails; unbind likewise fails and leaves the attach handle deallocation incomplete with ARQ_UNBIND_FAILED. These paths are unchanged by the patch, so bug #2174459\u0027s failure persists for MDEV/vGPU consumers.\n\n**Suggestion**:\nApply the same identity resolution in ExtARQ._allocate_attach_handle and the unbind path: resolve the device for the deployable (or ARQ device_rp) and target agent.create_vgpu_mdev/remove_vgpu_mdev with device.service_host, falling back to arq.hostname when absent. Extracting the small service_host-or-hostname resolution used in this patch (fpga_ext_arq.py:246-252) into a helper on Device would keep the rule in one place for all four callers.","commit_id":"d403874d68a94ac936d7f21e98bf96ce0b89fd0c"}]}
