chore(deps): bump github.com/hashicorp/consul/api from 1.33.7 to 1.34.2#406
chore(deps): bump github.com/hashicorp/consul/api from 1.33.7 to 1.34.2#406dependabot[bot] wants to merge 19 commits into
Conversation
886fe9c to
99cae81
Compare
Bumps [github.com/hashicorp/consul/api](https://github.com/hashicorp/consul) from 1.33.7 to 1.34.2. - [Release notes](https://github.com/hashicorp/consul/releases) - [Changelog](https://github.com/hashicorp/consul/blob/main/CHANGELOG.md) - [Commits](hashicorp/consul@api/v1.33.7...api/v1.34.2) --- updated-dependencies: - dependency-name: github.com/hashicorp/consul/api dependency-version: 1.34.2 dependency-type: direct:production update-type: version-update:semver-minor ... Signed-off-by: dependabot[bot] <support@github.com>
99cae81 to
ab84d77
Compare
Codecov Report❌ Patch coverage is
❌ Your patch check has failed because the patch coverage (0.00%) is below the target coverage (80.00%). You can increase the patch coverage or adjust the target coverage. ☔ View full report in Codecov by Sentry. 🚀 New features to boost your workflow:
|
ef34406 to
5da5977
Compare
The -coverpkg=./... flag combined with -race instruments all packages in every test binary, causing excessive memory usage. Serializing test package execution with -p=1 reduces peak memory to fit within CI runner limits (Go 1.26 race detector uses more memory than 1.25).
5da5977 to
dfa7383
Compare
Adds CoverEnv with GOMEMLIMIT=5500MiB applied when running coverage targets, using runWithEnv instead of run. Also ignores .antigravitycli/. Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
The race detector multiplies memory usage significantly. Combined with -coverpkg=./... and component/integration tests, this consistently exceeds the GitHub-hosted runner cgroup limit (exit code 137). Coverage measurement does not require race detection; -race stays in TestArgs for regular test runs. Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
-coverpkg=./... instruments all 22 packages into every test binary. Combined with component/integration tests that also spin up Docker containers, this exceeds the GitHub-hosted runner memory limit (exit code 137) even with -p=1 and no race detector. CoverUnit (unit tests only, no Docker) keeps -coverpkg=./... for full cross-package coverage. CoverAll uses per-package coverage instead, which keeps each binary's footprint small enough to survive. Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
The component test suite starts Kafka, MongoDB, Redis, Consul, Jaeger, AWSMock, MockServer, and a test service all at once. Combined these containers exhaust the 7 GB RAM on GitHub-hosted runners regardless of Go-level flags (-p=1, no -race, no -coverpkg). Use CoverUnit (unit tests only) for CI coverage; run CoverAll locally or on a self-hosted runner with sufficient memory. Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
The component test suite starts all containers simultaneously: Kafka + ZooKeeper (2× JVM), MockServer (JVM), MongoDB, Consul, Jaeger, AWSMock, and Redis. Without limits these collectively exceed the 7 GB available on GitHub-hosted runners. Changes: - Add MemoryMB field to SimpleContainerConfig; wired into HostConfig.Memory via the existing hcOpts callback in runContainer - Kafka: 512 MB cap + KAFKA_HEAP_OPTS=-Xmx256m; ZooKeeper: 256 MB + JVMFLAGS=-Xmx128m (keeps JVM within the cgroup limit) - MockServer: 256 MB + JAVA_OPTS=-Xmx128m; LOG_LEVEL lowered to WARN - MongoDB: 384 MB + --wiredTigerCacheSizeGB 0.25 - Consul: 128 MB, Jaeger: 256 MB, AWSMock: 256 MB, Redis: 64 MB Total containers: ~2.1 GB, leaving ~5 GB for Docker daemon, Go binary and OS. Restore CoverAll in CI now that containers fit within the limit. Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Two changes to reduce peak memory during component tests on CI runners: 1. StartComponents now uses SetLimit(3) so at most 3 components start simultaneously. Previously all 8 containers (including 3 JVMs) pulled images and initialized concurrently, causing a large memory spike. 2. JVM non-heap flags added to Kafka, ZooKeeper, and MockServer. -Xmx only limits the managed heap; Metaspace, CodeCache, and DirectMemory are unbounded without explicit caps. New flags: - Kafka: -XX:MaxMetaspaceSize=96m -XX:ReservedCodeCacheSize=32m -XX:MaxDirectMemorySize=64m - ZooKeeper: -XX:MaxMetaspaceSize=48m -XX:ReservedCodeCacheSize=16m -XX:MaxDirectMemorySize=32m - MockServer: -XX:MaxMetaspaceSize=64m -XX:ReservedCodeCacheSize=32m -XX:MaxDirectMemorySize=32m Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Two JVM processes (Kafka + ZooKeeper) were the dominant memory consumers
on CI runners. Even with aggressive heap flags (-Xmx256m/-Xmx128m) the
JVM non-heap (Metaspace, CodeCache, DirectMemory, native threads) added
another ~800 MB that Docker cgroup limits were not enforcing reliably.
Redpanda (redpandadata/redpanda) is a Kafka-compatible broker written in
C++. It runs as a single container with no ZooKeeper dependency and uses
~300 MB total, eliminating two JVM processes and roughly 2 GB of peak
memory pressure.
API is preserved: WithTopics("name:partitions:replication") still works.
NewComponent now extracts the KAFKA_CREATE_TOPICS env var (which Redpanda
does not understand) and creates topics via the sarama admin API in the
ready function instead.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
256M was below Seastar's minimum allocation threshold, causing Redpanda to abort with "Failed to allocate 64 bytes" before accepting connections. 512M allows Redpanda to start (~224MB actual RSS) and the Docker limit is raised to 640MB to provide headroom. Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
…ompilation RSS The kafka component now uses Redpanda's bundled rpk CLI (via InitExecCmd) for topic creation and the admin HTTP API for the ready check, removing the sarama import from production code. Sarama (35k lines) caused ~824MB peak RSS during cold compilation of the kafka sub-package on CI. Without sarama in kafka/, the cold compilation drops to ~216MB (GOGC=25) or ~309MB (GOGC=100). Adding GOGC=25 to CoverAll reduces peak RSS during each compilation pass by ~25-30%, trading extra GC cycles for lower memory headroom. Combined effect: peak compilation RSS for the component test binary falls from ~841MB to ~629MB on this machine; CI headroom should clear the 7GB runner limit. Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Pull request was closed
|
OK, I won't notify you again about this release, but will get in touch when a new version is available. If you'd rather skip all updates until the next major or minor version, let me know by commenting If you change your mind, just re-open this PR and I'll resolve any conflicts on it. |
Bumps github.com/hashicorp/consul/api from 1.33.7 to 1.34.2.
Commits
d739cbbupdating submodule versionf327aa6fix follow-redirects sec vuln (#23494) (#23496)f82280bGo version 1.26 (#23493) (#23497)b5716dcCVE suppression (#23478)a6fecaeBackport of Addition of json omitempty for ACLToken name field and its relate...1063b20Update the 2.0 major release version in the version file (#23480)aa36678removing CVE suppression (#23474)57e0866feat(multiport): add destination port support for multiport upstreams (#23449)f100b2cPost release activity 1.22.6 (#23462)fc8f069update: go version to 1.26.2 (#23394)