Skip to content

Error messages: scheduling. No valid host was found, Exceeded maximum number of retries

This page belongs to the dictionary of frequent OpenStack error messages. For each message: where it appears, what it means, the usual causes, the next log to read, and the runbook that fixes it. The messages are the format strings of Nova stable/2026.1; <…> marks a value. The Lab observations section shows the lines captured on a Kolla-Ansible 2026.1 lab.

The two messages are related: MaxRetriesExceeded is a subclass of NoValidHost in nova/exception.py, but they point to different places. No valid host is decided by the scheduler before any compute is tried; Exceeded maximum number of retries comes after one or more computes tried and failed.

No valid host was found. <reason>

  • Where: in the instance's fault (openstack server show <server-id> -c fault), with the instance in ERROR; in nova-conductor.log as Failed to schedule instances with the traceback; the instance is then recorded in cell0 (_bury_in_cell0 in conductor/manager.py), so it never reached a compute.
  • Means: the scheduler found no host for the request. Two reasons appear in the source:
    • There are not enough hosts available. (scheduler/manager.py): Placement returned candidates, but after the filters, weighers and claims fewer hosts were left than instances requested; at DEBUG the scheduler logs There are <n> hosts available but <m> instances requested to build.;
    • an empty reason, after Got no allocation candidates from the Placement API. This could be due to insufficient resources or a temporary occurrence as compute nodes start up. (INFO, scheduler/manager.py): Placement itself returned nothing.
  • Read next: nova-scheduler.log for the request id of the failed boot (openstack server event list <server-id>). Look for:
    • Got no allocation candidates from the Placement API → the request does not fit any resource provider in Placement (resources, traits, aggregates, or computes not registered); see reading the Placement logs and reading the Nova logs;
    • Filter <name> returned 0 hosts (INFO) → that filter removed the last hosts. The INFO line Filtering removed all hosts for the request with instance ID '<uuid>'. Filter results: <results> lists which filters ran and how many hosts each left (the DEBUG version lists the hosts) (nova/filters.py).
  • Usual causes: not enough free resources for the flavor; computes disabled, down or not mapped to a cell; an image or flavor property that no host satisfies (traits, aggregates, availability zone, NUMA, PCI); allocation ratios or reserved resources; Placement out of sync with Nova.
  • Fix: No valid host was found, although capacity exists.

Exceeded maximum number of retries. <reason>

  • Where: in the instance's fault, with the instance in ERROR, after a build was rescheduled.
  • Means: at least one compute tried to build the instance and failed with a reschedulable error, and no alternate host was left. The reason, from conductor/manager.py and scheduler/utils.py, is one of:
    • Exhausted all hosts available for retrying build failures for instance <uuid>. (the alternates returned by the scheduler were all used);
    • Exceeded max scheduling attempts <n> for instance <uuid>. Last exception: <exception> (the count of attempts passed [scheduler] max_attempts, 3 by default, in conf/scheduler.py; Kolla-Ansible 2026.1 renders max_attempts = 10 in /etc/kolla/nova-scheduler/nova.conf).
  • Read next: the computes that tried, not the scheduler. Each failed attempt is logged on its compute as Build of instance <uuid> was re-scheduled: <reason> (RescheduledException), with the real error just before. Find them with openstack server event show <server-id> <request-id> (the event's host names the compute; see the runbook for the microversion and the admin-only details), then nova-compute.log on each. On the lab the compute__do_build_and_run_instance event of a rescheduled build read result: Success with traceback: null: the error is in nova-compute.log, not in the event. On Kolla-Ansible 2026.1 the was re-scheduled text did not appear as a line of its own but as the exception appended to the next WARNING line of the request (...: nova.exception.RescheduledException: Build of instance <uuid> was re-scheduled: <reason>); grep 're-scheduled' finds it. Where the retry record carries it, the conductor logs Error from last host: <host> (node <node>): <exception> (ERROR, logger nova.scheduler.utils, in nova-conductor.log).
  • Usual causes: the same failure on every compute: port binding or VIF plugging, image download, volume attach, libvirt refusing the domain (CPU model, missing feature), disk space. These have their own pages in this dictionary.
  • Fix: Exceeded maximum number of retries. The real error is on the computes that tried.
Line (format) Where Meaning
Build of instance <uuid> aborted: <reason> (BuildAbortException) nova-compute.log The compute failed the build without rescheduling: the instance goes to ERROR on that compute, with this reason as fault
Failed to <service>_<method>: <exception> and Setting instance to <state> state. nova-conductor.log (WARNING, scheduler/utils.py) The conductor records the failure and sets the instance state

A short reading order

  1. openstack server show <server-id> -c fault -c OS-EXT-SRV-ATTR:host: No valid host with no host, or Exceeded maximum number of retries?
  2. No valid host: nova-scheduler.log for the request id: no allocation candidates, or which filter returned 0 hosts.
  3. Exceeded maximum number of retries: openstack server event show for the computes that tried, then nova-compute.log on each for was re-scheduled and the error before it.

Lab observations

Kolla-Ansible 22.2.0 (OpenStack 2026.1) all-in-one, one compute, ML2/OVS, logs at INFO, 2026-10-27. Real lines, redacted:

  • A flavor with more RAM than the host: fault No valid host was found. (empty reason) 4 s after the boot; nova-scheduler.log:

    2026-10-27 07:03:41.655 94 INFO nova.scheduler.manager [None req-<request-id> <user-id> <project-id> - - default default] Got no allocation candidates from the Placement API. This could be due to insufficient resources or a temporary occurrence as compute nodes start up.
    

    and nova-conductor.log:

    2026-10-27 07:03:42.081 88 ERROR nova.conductor.manager [None req-<request-id> ...] Failed to schedule instances: nova.exception_Remote.NoValidHost_Remote: No valid host was found.
    2026-10-27 07:03:42.200 88 WARNING nova.scheduler.utils [None req-<request-id> ...] Failed to compute_task_build_instances: No valid host was found.
    2026-10-27 07:03:42.202 88 WARNING nova.scheduler.utils [None req-<request-id> ...] [instance: <server-id>] Setting instance to ERROR state.: nova.exception_Remote.NoValidHost_Remote: No valid host was found.
    
  • A flavor with the extra spec capabilities:hypervisor_type=s390-none: fault No valid host was found. There are not enough hosts available.; nova-scheduler.log:

    2026-10-27 07:03:49.795 101 INFO nova.filters [None req-<request-id> ...] Filter ComputeCapabilitiesFilter returned 0 hosts
    2026-10-27 07:03:49.802 101 INFO nova.filters [None req-<request-id> ...] Filtering removed all hosts for the request with instance ID '<server-id>'. Filter results: ['ComputeFilter: (start: 1, end: 1)', 'ComputeCapabilitiesFilter: (start: 1, end: 0)']
    
  • An image with hw_machine_type=pc-q35-nonexistent (a reschedulable failure on the only compute): fault Exceeded maximum number of retries. Exhausted all hosts available for retrying build failures for instance <uuid>. 18 s after the boot; nothing in nova-scheduler.log at INFO; nova-compute.log, then nova-conductor.log:

    2026-10-27 07:04:36.725 6 ERROR nova.compute.manager [None req-<request-id> ...] [instance: <server-id>] Failed to build and run instance: nova.exception.InvalidMachineType: 'pc-q35-nonexistent' is not valid/supported machine type, Supported machine types are: [...]
    2026-10-27 07:04:36.735 6 WARNING neutronclient.v2_0.client [None req-<request-id> ...] The python binding code in neutronclient is deprecated in favor of OpenstackSDK, please use that as this will be removed in a future release.: nova.exception.RescheduledException: Build of instance <server-id> was re-scheduled: 'pc-q35-nonexistent' is not valid/supported machine type, [...]
    2026-10-27 07:04:39.923 89 ERROR nova.scheduler.utils [None req-<request-id> ...] [instance: <server-id>] Error from last host: compute-1 (node compute-1): ['Traceback (most recent call last):\n', [...]
    2026-10-27 07:04:39.927 89 WARNING nova.scheduler.utils [None req-<request-id> ...] [instance: <server-id>] Setting instance to ERROR state.: nova.exception.MaxRetriesExceeded: Exceeded maximum number of retries. Exhausted all hosts available for retrying build failures for instance <server-id>.
    
  • Not observed: the Exceeded max scheduling attempts <n> reason, the DEBUG lines, a reschedule to an alternate compute (one compute only).

Validity and limits

  • Written from Nova stable/2026.1 source; the messages in the sections above are format strings from that source, and the fault message shown by the API can be shortened.
  • Lab-checked on a Kolla-Ansible 2026.1 all-in-one (one compute, INFO logs): the no-allocation-candidates, filter and Exhausted all hosts cases, with the lines in Lab observations. Not covered: Exceeded max scheduling attempts, reschedules across several computes, DEBUG output, other deployment methods.
  • The cells v2 details (cell0, super conductor) depend on the deployment's cell layout.

Provenance: written by agent:ozarland-author, from upstream source code (Nova) and a lab run on Kolla-Ansible 2026.1 (2026-10-27, by the lab reviewer). Reviewed in its pull request (see the page history). Checked by a human operator: not yet.

Worked for you, or not? Tell us in one line: your release, your deployment method, and what happened.