Select READY images while preserving architecture filters, newest-first
ordering and explicit image overrides. Use a singleton status list to
avoid nested Compute API filter expressions.
Signed-off-by: thc1006 <84045975+thc1006@users.noreply.github.com>
The additional commands (mounter, gcp-credentials-provider)
were needed for E2E node testing. Previously, the code got duplicated
for linking into e2e_node.test. Now the old code gets removed.
gsutil will no longer ship with the Google Cloud CLI starting March 2027.
kube-up now stages release tars with gcloud storage and no longer
requires gsutil in PATH. The md5 check uses "gcloud storage hash
--format=value(md5_hash)", which prints the bare hex digest, so the
grep/awk/sed parsing of gsutil output is gone.
stage-upload.sh for the gci mounter switches to "gcloud storage cp".
describe/delete exit 1 for both 404s and transient errors, so a single
flake was treated as "already deleted". On a failed delete, check with
list instead: empty output means gone, an error means retry.
Update konnectivity network proxy to v0.37.0
Include CVE fixes.
Decoupled send/recv in data-plane to control-plane in agent.
Improve connection selection from server.
Unified HTTP-CONNECT and GRPC packet handling in server.
Improved HTTP-CONNECT handling for slow connection.
Reject malformed certificate blocks in CA bundles.
Added connection_duration_seconds histogram.
Added exported DialFailureReason on setup.
Added reuable tunnel.
Fix CFSSL build on mac.
Fix agent drain signal blocked by stream RECV.
Made CloseHTTP idempotent.
Fix removeEstablishedForBackendConn from retiring entire backend.
Fixed race condition on response through agent.
Made lease labels a real label selector.
Restored non-graceful shutdown behavior.
Update dependencies ANP dependencies.
Update ANP client dependencies to match K/K 1.34.
gcloud compute images list --filter="family:X" returns every
non-deprecated image in the family. On 2026-07-31 cos-cloud briefly
carried two active images in cos-129-lts (cos-129-19506-299-60 and
-61, later rolled back), so the newline-joined result produced an
invalid sourceImage URL and every GCE e2e job failed cluster bring-up
with "Failed to create master instance due to non-retryable error".
Sort by creationTimestamp and take the newest image, matching the
image-family head semantics.
Update the default COS image family from cos-125-lts to cos-129-lts
which ships containerd v2.2.5 and runc v1.4.0. Containerd v2.2+
implements ListPodSandboxMetrics with per-container metrics, which is
needed to test the CRI stats path. Also update the COS GPU installer
from v2.5.8 to v2.6.8 for M129 kernel compatibility.
In E2E reboot tests that simulate crashes (unclean shutdowns), the
systemd journal logs on the writeable partition can become corrupted
or truncated.
On the subsequent boot, the GCE node configuration scripts
(`configure.sh` and `configure-helper.sh`) attempt to read the boot
history using `journalctl --list-boots` to initialize telemetry logging.
Due to the journal corruption, `journalctl` fails with "No data
available". Because the scripts run with `set -o errexit` enabled, this
telemetry failure crashes the entire node setup early, preventing the
node from becoming Ready.
This fix replaces the dependency on `journalctl` for boot ID detection
with reading directly from `/proc/sys/kernel/random/boot_id`, which is
in-memory and immune to disk/journal corruption.
Refs #139444
Signed-off-by: Huy Pham <huypham@google.com>
Update konnectivity network proxy to v0.36.0.
Includes CVE fixes. Adds pending backend request timeout option.
Enforce server-ca-cert in TCP mode validation.
add --tls-min-version flag to proxy server
Remove the remaining node-problem-detector references across the cluster/
directory that were not covered by the initial cleanup:
- Remove ENABLE_NODE_PROBLEM_DETECTOR, NODE_PROBLEM_DETECTOR_VERSION,
NODE_PROBLEM_DETECTOR_TAR_HASH, NODE_PROBLEM_DETECTOR_RELEASE_PATH and
NODE_PROBLEM_DETECTOR_CUSTOM_FLAGS exports from config-default.sh and
config-test.sh.
- Remove WINDOWS_ENABLE_NODE_PROBLEM_DETECTOR and
WINDOWS_NODE_PROBLEM_DETECTOR_CUSTOM_FLAGS from config-default.sh and
config-test.sh.
- Remove WINDOWS_NODEPROBLEMDETECTOR_KUBECONFIG_FILE from config-common.sh.
- Remove create-node-problem-detector-kubeconfig and
create-node-problem-detector-kubeconfig-from-kubelet functions along
with all NPD token/kubeconfig setup blocks from configure-helper.sh.
- Remove NPD installation and configuration steps (install-npd,
configure-npd, start-node-problem-detector) from upgrade.sh and util.sh.
- Remove DownloadAndInstall-NodeProblemDetector,
Create-NodeProblemDetectorKubeConfig and Configure-NodeProblemDetector
from cluster/gce/windows/k8s-node-setup.psm1 and their call sites in
configure.ps1.
- Remove the node-problem-detector systemd source block from the fluentd
configmaps and the node-problem-detector.log entry from log-dump.sh.
- Remove npd and npdSA test variables and their use from audit_policy_test.go.
Co-authored-by: Jordan Liggitt <liggitt@google.com>
Signed-off-by: Humble Devassy Chirammal <humble.devassy@gmail.com>
Update node-problem-detector from v1.34.0 to v1.35.2 and remove all
related addon manifests and install logic that is no longer needed:
- Update version in build/dependencies.yaml, test/e2e_node/image_list.go
and test/kubemark/resources/hollow-node_template.yaml.
- Remove cluster/addons/node-problem-detector/ entirely. No e2e tests
depend on these manifests: e2e_node tests create NPD pods inline and
GCE standalone mode runs NPD as a systemd service.
- Remove install-node-problem-detector function and DEFAULT_NPD_* vars
from cluster/gce/gci/configure.sh along with the conditional that
invoked it, since NPD is no longer installed as a standalone binary
via this script.
- Remove the setup-addon-manifests calls for node-problem-detector from
cluster/gce/gci/configure-helper.sh since the source directory no
longer exists.
- Remove stale refPaths in build/dependencies.yaml that pointed to the
deleted addon files.
Signed-off-by: Humble Devassy Chirammal <humble.devassy@gmail.com>
kubernetes-prow[bot]