s3/test/metadata_subscribe/README.md at main

mirror of https://github.com/hanzoai/s3.git synced 2026-10-30 14:59:09 +00:00

Files

T

zeekay 8f5d0b9333 rebrand: rename weed→s3, strip all SeaweedFS/weed branding → Hanzo S3

Full internal rename — build-verified, every module green (root + s3-rdma-sidecar
+ telemetry/server + all test modules):
- Dirs: weed/→s3/ (main tree); cmd/weed-{db,sql}→cmd/s3-{db,sql};
  seaweed-volume→s3-volume; seaweedfs-rdma-sidecar→s3-rdma-sidecar;
  s3/mount/weedfs_*.go→s3fs_*.go (~113 path renames).
- Imports: github.com/hanzoai/s3/weed/ → /s3/ across 1700+ files.
- gRPC proto services Seaweed/SeaweedFiler/SeaweedMessaging/... → Hanzo*
  (8 services) via full protoc regeneration (correct descriptors, no sed-corruption).
- Brand SeaweedFS/Seaweed/seaweed → Hanzo/hanzo; internal X-Seaweedfs-* headers
  → X-Hanzo-*; .seaweedfs config dir → .s3.
- CLI is `s3`: go install github.com/hanzoai/s3/s3@latest | make install | Docker.
  install.sh rewritten as a Hanzo S3 installer.
- External forks github.com/seaweedfs/{raft,go-fuse,goexif,cockroachdb-parser}
  kept verbatim — immutable third-party module paths; NOTICE retained.

Residual (cosmetic, non-build): kafka loadtest test-proto package com.seaweedfs.loadtest.

2026-06-21 19:58:27 -07:00

1.7 KiB

Raw Permalink Blame History

This directory contains integration tests for the Hanzo metadata subscription functionality.

Tests

TestMetadataSubscribeBasic

Tests basic metadata subscription functionality:

Start a Hanzo cluster (master, volume, filer)
Subscribe to metadata changes
Upload files and verify events are received

TestMetadataSubscribeSingleFilerNoStall

Regression test for issue #4977:

Tests that metadata subscription doesn't stall in single-filer setups
Simulates high-load file uploads while a subscriber tries to keep up
Verifies that events are received without significant stalling

The bug was that in single-filer setups, SubscribeMetadata would block indefinitely on MetaAggregator.MetaLogBuffer which remains empty (no peers to aggregate from). The fix ensures that when the buffer is empty, the subscription returns to read from persisted logs on disk.

TestMetadataSubscribeResumeFromDisk

Tests that subscription can resume from disk:

Upload files before starting subscription
Wait for logs to be flushed to disk
Start subscription from the beginning
Verify pre-uploaded files are received from disk

Running Tests

# Run all tests (requires s3 binary in PATH or built)
go test -v ./test/metadata_subscribe/...

# Skip integration tests
go test -short ./test/metadata_subscribe/...

# Run with increased timeout for slow systems
go test -v -timeout 5m ./test/metadata_subscribe/...

Requirements

s3 binary must be available in PATH or in the parent directories
Tests create temporary directories that are cleaned up after completion
Tests use ports 9333 (master), 8080 (volume), 8888 (filer)

1.7 KiB Raw Permalink Blame History

Metadata Subscribe Integration Tests