Remove pending vm handling mechanism from scheduler + fixed issue that update VM's hostname even on migration failure

scripts/ucloud

@@ -24,7 +24,7 @@ sys.excepthook = exception_hook
 if __name__ == '__main__':
     # Setting up root logger
     logger = logging.getLogger()
-    logger.setLevel(logging.INFO)
+    logger.setLevel(logging.DEBUG)
 
     syslog_handler = SysLogHandler(address='/dev/log')
     syslog_handler.setLevel(logging.DEBUG)

ucloud/api/main.py

@@ -567,7 +567,7 @@ def main():
             settings["etcd"]["image_store_prefix"], value_in_json=True
         )
     )
-    if len(image_stores) == 0:
+    if not image_stores:
         data = {
             "is_public": True,
             "type": "ceph",
@@ -583,7 +583,7 @@ def main():
             json.dumps(data),
         )
 
-    app.run(host="::", debug=True)
+    app.run(host="::", debug=False)
 
 
 if __name__ == "__main__":

ucloud/host/main.py

@@ -28,10 +28,8 @@ def maintenance(host):
     vmm = VMM()
     running_vms = vmm.discover()
     for vm_uuid in running_vms:
-        if (
+        if vmm.is_running(vm_uuid) and vmm.get_status(vm_uuid) == "running":
-            vmm.is_running(vm_uuid)
+            logger.debug('VM {} is running on {}'.format(vm_uuid, host))
-            and vmm.get_status(vm_uuid) == "running"
-        ):
             vm = shared.vm_pool.get(
                 join_path(settings["etcd"]["vm_prefix"], vm_uuid)
             )
@@ -43,32 +41,18 @@ def maintenance(host):
 
 def main(hostname):
     host_pool = shared.host_pool
-    host = next(
+    host = next(filter(lambda h: h.hostname == hostname, host_pool.hosts), None)
-        filter(lambda h: h.hostname == hostname, host_pool.hosts), None
+    assert host is not None, "No such host with name = {}".format(hostname)
-    )
-    assert host is not None, "No such host with name = {}".format(
-        hostname
-    )
 
     try:
-        heartbeat_updating_process = mp.Process(
+        heartbeat_updating_process = mp.Process(target=update_heartbeat, args=(hostname,))
-            target=update_heartbeat, args=(hostname,)
-        )
         heartbeat_updating_process.start()
     except Exception as e:
-        raise Exception(
+        raise Exception("ucloud-host heartbeat updating mechanism is not working") from e
-            "ucloud-host heartbeat updating mechanism is not working"
-        ) from e
 
     for events_iterator in [
-        shared.etcd_client.get_prefix(
+        shared.etcd_client.get_prefix(settings["etcd"]["request_prefix"], value_in_json=True),
-            settings["etcd"]["request_prefix"], value_in_json=True
+        shared.etcd_client.watch_prefix(settings["etcd"]["request_prefix"], timeout=10, value_in_json=True)
-        ),
-        shared.etcd_client.watch_prefix(
-            settings["etcd"]["request_prefix"],
-            timeout=10,
-            value_in_json=True,
-        ),
     ]:
         for request_event in events_iterator:
             request_event = RequestEntry(request_event)
@@ -76,20 +60,13 @@ def main(hostname):
             if request_event.type == "TIMEOUT":
                 maintenance(host.key)
 
-            if request_event.hostname == host.key:
+            elif request_event.hostname == host.key:
-                logger.debug("VM Request: %s", request_event)
+                logger.debug("VM Request: %s on Host %s", request_event, host.hostname)
-
+                shared.request_pool.client.client.delete(request_event.key)
-                shared.request_pool.client.client.delete(
-                    request_event.key
-                )
                 vm_entry = shared.etcd_client.get(
-                    join_path(
+                    join_path(settings["etcd"]["vm_prefix"], request_event.uuid)
-                        settings["etcd"]["vm_prefix"],
-                        request_event.uuid,
                 )
-                )
+                logger.debug("VM hostname: {}".format(vm_entry.value))
-
-                if vm_entry:
                 vm = virtualmachine.VM(vm_entry)
                 if request_event.type == RequestType.StartVM:
                     vm.start()
@@ -100,28 +77,18 @@ def main(hostname):
                 elif request_event.type == RequestType.DeleteVM:
                     vm.delete()
 
-                    elif (
+                elif request_event.type == RequestType.InitVMMigration:
-                        request_event.type
-                        == RequestType.InitVMMigration
-                    ):
                     vm.start(destination_host_key=host.key)
 
                 elif request_event.type == RequestType.TransferVM:
-                        host = host_pool.get(
+                    destination_host = host_pool.get(request_event.destination_host_key)
-                            request_event.destination_host_key
+                    if destination_host:
-                        )
-                        if host:
                         vm.migrate(
-                                destination_host=host.hostname,
+                            destination_host=destination_host.hostname,
                             destination_sock_path=request_event.destination_sock_path,
                         )
                     else:
-                            logger.error(
+                        logger.error("Host %s not found!", request_event.destination_host_key)
-                                "Host %s not found!",
-                                request_event.destination_host_key,
-                            )
-                else:
-                    logger.info("VM Entry missing")
 
 
 if __name__ == "__main__":

ucloud/host/virtualmachine.py

@@ -38,13 +38,14 @@ class VM:
         else:
             self.uuid = vm_entry.key.split("/")[-1]
             self.host_key = self.vm["hostname"]
+        logger.debug('VM Hostname {}'.format(self.host_key))
 
     def get_qemu_args(self):
         command = (
-            "-name {owner}_{name}"
+            "-drive file={file},format=raw,if=virtio,cache=none"
-            " -drive file={file},format=raw,if=virtio,cache=none"
             " -device virtio-rng-pci"
             " -m {memory} -smp cores={cores},threads={threads}"
+            " -name {owner}_{name}"
         ).format(
             owner=self.vm["owner"],
             name=self.vm["name"],
@@ -67,11 +68,7 @@ class VM:
         except Exception as err:
             declare_stopped(self.vm)
             self.vm["log"].append("Cannot Setup Network Properly")
-            logger.error(
+            logger.error("Cannot Setup Network Properly for vm %s", self.uuid, exc_info=err)
-                "Cannot Setup Network Properly for vm %s",
-                self.uuid,
-                exc_info=err,
-            )
         else:
             self.vmm.start(
                 uuid=self.uuid,
@@ -81,6 +78,7 @@ class VM:
             )
 
             status = self.vmm.get_status(self.uuid)
+            logger.debug('VM {} status is {}'.format(self.uuid, status))
             if status == "running":
                 self.vm["status"] = VMStatus.running
                 self.vm["vnc_socket"] = self.vmm.get_vnc(self.uuid)
@@ -99,7 +97,7 @@ class VM:
             else:
                 self.stop()
                 declare_stopped(self.vm)
-
+        logger.debug('VM {} has hostname {}'.format(self.uuid, self.vm['hostname']))
         self.sync()
 
     def stop(self):

ucloud/scheduler/main.py

@@ -17,8 +17,6 @@ from . import logger
 
 
 def main():
-    pending_vms = []
-
     for request_iterator in [
         shared.etcd_client.get_prefix(
             settings["etcd"]["request_prefix"], value_in_json=True
@@ -44,24 +42,8 @@ def main():
                     logger.debug("Dead hosts: %s", dead_hosts)
                     dead_host_mitigation(dead_hosts)
 
-                # If there are VMs that weren't assigned a host
-                # because there wasn't a host available which
-                # meets requirement of that VM then we would
-                # create a new ScheduleVM request for that VM
-                # on our behalf.
-                while pending_vms:
-                    pending_vm_entry = pending_vms.pop()
-                    r = RequestEntry.from_scratch(
-                        type="ScheduleVM",
-                        uuid=pending_vm_entry.uuid,
-                        hostname=pending_vm_entry.hostname,
-                        request_prefix=settings["etcd"][
-                            "request_prefix"
-                        ],
-                    )
-                    shared.request_pool.put(r)
-
             elif request_entry.type == RequestType.ScheduleVM:
+                print(request_event.value)
                 logger.debug(
                     "%s, %s", request_entry.key, request_entry.value
                 )
@@ -86,7 +68,6 @@ def main():
                     )
                     shared.vm_pool.put(vm_entry)
 
-                    pending_vms.append(vm_entry)
                     logger.info("No Resource Left. Emailing admin....")