Tags

v0.30.0

Lots of internal changes, mostly tech debt and reliability improvements,
but some bugfixes. Also of note: Kernel upgraded from 6.1.63 -> 6.1.92.

Features:

- api,agent/core: Make required ScalingConfig fields optional per-VM (#948)
- neonvm-controller: replace failing reconciliation with per-VM failure interval  (#949)

No breaking changes.

No protocol changes.

Fixes:

- neonvm-controller: Rework zap setup & disable sampling (#947)
- neonvm-controller: Fix k3s cluster detection (#956)

Other changes:

- plugin: Remove vmStore from AutoscaleEnforcer (#904)
- plugin: Handle Buffer in reserveResources() (#904)
- plugin: Replace readClusterState with existing watch events (#904)
- neonvm-controller: rename VirtualMachine -> VM (#942)
- pg16-disk-test: add mkfifo (#944)
- Use github.com/samber/lo.ToPtr everywhere (#946)
- neonvm: Switch to kernel 6.1.92 (#950)

Upgrade path from v0.29.x:

- No ordering requirements.

v0.29.3 but with 1.26 k8s API

v0.29.3

Single change, just #937, changing neonvm-controller metrics to
differentiate in the "currently failing reconcile" count between objects
that were failing to reconcile due to conflicts vs objects that were
failing from some other errors.

v0.29.2

Single change, aiming to ease rollout for S3 billing chang:
e3b2cec agent: don't check S3.Endpoint for emptyness (#934)

v0.29.1

A hotfix after v0.29.0, containing a version bump, to fix S3 access:
d932bc5 deps: upgrade aws/config and s3 packages (#933)

v0.29.0

No known breaking changes.

New features:
9fbb531 runner: allow coredump collection (#931)
3b08db3 neonvm: metrics for failing reconcile (#920)
da5dbf2 agent: push metrics to S3 (#898)

Bug fixes:
669bb52 agent: don't treat downscale denies as failed request (#927)

Deployment changes:
8601caf neonvm-runner-image-loader: Faster rollout (#923)

Deps updates:
56a38a4 build(deps): bump google.golang.org/protobuf from 1.30.0 to 1.33.0 (#858)
5b2b27e Update Rust: 1.74 -> 1.78 (#925)
fa2c634 build(deps): bump golang.org/x/net from 0.17.0 to 0.23.0

Developer workflow changes:
774a6d5 Generate manifests via `make generate`, remove `make manifests` (#919)
4d83eb6 github: Fix improvement template typo: facting => facing (#917)
eff79b6 github: mv improvement_template.md => improvement-template.md (#917)

v0.28.1 but with 1.26 k8s API

v0.28.1

This release fixes a critical issue in the backwards compatibility of
v0.28.0 (and subsequent patches).

In particular, it changes the type of the swap setting back to
*resource.Quantity, and introduces a new field for *SwapInfo.

See #912 for more info.

v0.28.0-patch2

Another quick follow-up, fixing HTTP 400s for old autoscaler-agent
instances.

See #910 for more.

v0.28.0-patch1

Tiny release, containing the fix for a panic in the scheduler plugin's
handling of autoscaler-agent requests.

See #909 for more details.

v0.28.0

Large release after a while - lots of PRs without external changes, and
then a hanful all at once.

Features:

- plugin: Make auto- live migration opt-in (#781)
- plugin: Count shouldDeny from Reserve (#891)
  - new metric: autoscaling_plugin_reserve_should_deny_total
- agent: Track more liveness in vm-stuck metrics (#855)
- plugin: Add eventQueueSet metrics (#875)
  - new metric: autoscaling_plugin_eventqueue_depth
  - new metric: autoscaling_plugin_eventqueue_adds_total
  - new metric: autoscaling_plugin_eventqueue_duration_seconds
- neonvm: Support resizeable swap via sized mkswap (#887)

Breaking changes:

- neonvm: Remove support for pre-v0.7.0 neonvm-runner versions (#836)
- agent/billing: Make clients generic and allow >1 (#890)
  - Config format changed. See PR for details.
- agent: Make billing config non-optional (#890)
- neonvm: Changed type of VirtualMachine.Spec.Guest.Settings.Swap

Protocol changes:

- agent,plugin: Remove unused metrics from api.Metrics (#750)

Fixes:

- Sync versioning in VERSIONING.md and plugin (#894)
  - DISALLOWS versions <v3.0, which should have already been the case.
- agent: Add missing validationf or .metrics.requestTimeoutSeconds
  (#892)

Other changes:

- api/vminfo: Move flags & scaling config to new Config field (#781)
- plugin: Use api.VmConfig for per-VM settings (#781)
- plugin: Don't fail in Reserve() (#891)
- neonvm-runner: Log parallel errors before returning (#893)
- agent,api: Remove custom Format()s for complex types (#902)

Upgrade path from v0.27.0:

- neonvm MUST be upgraded before any Go client libraries, if the client
  does any interactions with VMs with swap enabled.
- the scheduler MUST be upgraded before the autoscaler-agents.

v0.27.0

Incident followups:
 - c659485 plugin: construct VmInfo from pod, not vm (#865)
 - 63e618d plugin: Increase eventQueueWorkers: 16 -> 64 (#874)
 - 1fa43fa neonvm: Add neonvm-runner image loader (#857)
 - 6878837 plugin: Add pod desired AZ to plugin method metrics (#870)

Other changes:
- 63e1d72 runner: refactor to make some steps parallel (#864)
- e7f8039 Release workflow: add neonvm-runner-image-loader (#876)

Upgrade path from v0.26.2:
 - Need to deploy neonvm-controller before scheduler to wait for
 propagation of annotations

v0.26.2

Small release containing only a few PRs - releasing now to get a
critical scheduler fix out (see #851 for more).

Fixes:

- plugin: Use multiple event handler threads (#854)

Other changes:

- neonvm-controller: Enable seccomp profile (#833)
- neonvm: Make .status.restartCount not a pointer (#845)
- neonvm: Remove controller metrics auth proxy (#825)

Upgrade path from v0.26.x:

- No ordering requirements.

v0.26.1

Patch release with only the contents of:

 - agent/dispatcher: Fix inbound monitor msg metrics (#838)

Made as a part of autoscaling release training.

v0.26.0

Fixes:
- neonvm-controller: Catch Reconcile() panics (#829)

Other changes:
- upgrade test vm image to pg 16 (#810)
- neonvm: increase neonvm-controller replicas (#823)
- neonvm: log VM name in the neonvm webhook manager (#816)
- neonvm: Cleanup kustomize (#824)

Upgrade path from v0.25.0:

- No ordering requirements.

v0.25.1

Patch release with only the contents of:

- plugin: Remove client_addr metrics label (#815)

Refer to the PR for more info.

v0.25.0

Even mix of fixes and features, centered around improving stability and
reliability.

Features:

- neonvm: Add support for having a swap disk in VMs (#801)
- neonvm: Add InitScript to spec (#782)

No breaking changes.

No protocol changes.

Fixes:

- plugin: Fix k8s status message if pod scheduling fails (#804)
- neonvm-controller: Fix field name typo in webhook validation (#793)
- neonvm-runner: log forward: Wait for QEMU startup and prevent panic (#811)

Other changes:

- plugin: Handle panics from agent requests (#785)
- neonvm-controller: Bump CPU limits 4 -> 8 (#805)
- neonvm-controller: Enable container-mgr by default (#803)
- neonvm: Merge init-roodisk and sysctl containers (#769)
- neonvm: Tune QEMU disk cache settings (#776)
- neonvm-controller: Add reconcile metrics debug endpoint (#798)

Upgrade path from v0.24.0:

- No ordering requirements.

v0.24.0

Large release - lots of fixes, with a side of new features thrown in to
spice it up. This is actually a re-attempt of the same release, due to
issues and noticing a critical bug, that have now been fixed.

Features:

- neonvm-controller: Make max concurrency configurable via CLI (#773)
  - This was previously released as v0.23.2.
- neonvm: Add .status.restartCount (#754)
- neonvm-controller: VM startup metrics (#774)

No breaking changes.

No protocol changes.

Fixes:

- neonvm-controller: Fix overwriting runner version (#753)
- neonvm-runner: Don't log i/o timeout (#768)
- neonvm: Use crictl to change container CPU, ditch cgroup (#738)
  - This is a BIG change! It's disabled by default, can be enabled with
    a neonvm-controller flag.
- neonvm-runner: Fix "dnsmasq: failed to create inotify .." errors (#786)
- agent: Fix unregistered metric lastSendDuration (#787)
- neonvm-controller: Don't retry update on conflict (#796)
- neonvm-controller: Update VM's .Status.PodName immediately on API server (#797)

Other changes:

- neonvm-runner: Skip QEMU powerdown if already exited (#526)
- plugin: Unify reserve and unreserve logic (#666)
- neonvm: Use container statuses, not pod phase, to trigger restart (#749)
- neonvm-controlller: Use --concurrency-limit=128 (#783)

Upgrade path from v0.23.x:

- No ordering requirements.

v0.23.2

Backport of #773 to allow configuring max neonvm-controller max
concurrency.

Now that we have neonvm-controller metrics, we discovered that a couple
regions were sometimes saturating the concurrency limit for extended
periods of time.

Refer to #773 for more details.