Files
seaweedfs/test/erasure_coding
Chris Lu 5d8a463b3e test/ec: fix EC interruption matrix slot exhaustion (#11295)
* test/ec: fix EC interruption matrix slot exhaustion

The EC integration test cluster (test/erasure_coding/chaos_lifecycle_test.go)
configured each disk with -max 4 and the seedAndSpread spread loop fired
volume.grow -count 4 every 2 s with no per-server cap. Because the master
topology lags the volume.grow writes, the loop re-fired before the prior
grow was visible, over-filling disks to capacity. A full disk leaves zero
free EC shard slots (failing the cluster-wide capacity check with "no
free ec shard slots") and drops the source disk below the encode's
FreeVolumeCount >= 2 health check (failing with "no healthy replicas"),
which aborted ec.encode before any phase marker printed and made every
encode scenario in TestECInterruptionMatrix fail.

Three changes to the test cluster:

1. Raise -max from 4 to 8 per disk so the source disk always retains
   FreeVolumeCount >= 2 for ec.encode's 14-shard generation (2 volume-slot
   equivalents) even after the spread loop and multiple encodes.

2. Switch the spread loop from -count 4 to -count 1 so each grow lands
   exactly one volume on the volume server's least-loaded disk, giving
   deterministic cross-disk spreading instead of relying on a single
   multi-volume grow to fan out.

3. Cap grows per server at 4 so heartbeat lag cannot run away and
   over-fill disks before the master registers the prior grow.

4. Pass -minFreeSpace 0 so the test is not falsely gated by the physical
   disk's free-space percentage on the host running CI (the EC shard slot
   calculation separately enforces a 90 % disk-usage cap via
   balancer.DiskTooFullAfter, which already guards against an over-set
   maxVolumeCount on a physically full disk).

Verified locally by running TestECInterruptionMatrix twice (all encode,
decode, and balance scenarios pass, including the previously failing
encode@Deletingoriginalvolumes).

* test/ec: only count successful grows toward the spread cap

A failed volume.grow (e.g. a transient collectTopologyInfo or VolumeGrow
RPC error) would otherwise consume one of the four permitted attempts
without creating any volume, exhausting the retry budget and leaving the
loop to only poll until the Eventually timeout. Increment the per-server
counter only when commandGrow.Do returns nil.
2026-09-13 14:41:45 -07:00
..

Erasure Coding Integration Tests

This directory contains integration tests for the EC (Erasure Coding) encoding volume location timing bug fix.

The Bug

The bug caused double storage usage during EC encoding because:

  1. Silent failure: Functions returned nil instead of proper error messages
  2. Timing race condition: Volume locations were collected AFTER EC encoding when master metadata was already updated
  3. Missing cleanup: Original volumes weren't being deleted after EC encoding

This resulted in both original .dat files AND EC .ec00-.ec13 files coexisting, effectively doubling storage usage.

The Fix

The fix addresses all three issues:

  1. Fixed silent failures: Updated doDeleteVolumes() and doEcEncode() to return proper errors
  2. Fixed timing race condition: Created doDeleteVolumesWithLocations() that uses pre-collected volume locations
  3. Enhanced cleanup: Volume locations are now collected BEFORE EC encoding, preventing the race condition

Integration Tests

TestECEncodingVolumeLocationTimingBug

The main integration test that:

  • Simulates master timing race condition: Tests what happens when volume locations are read from master AFTER EC encoding has updated the metadata
  • Verifies fix effectiveness: Checks for the "Collecting volume locations...before EC encoding" message that proves the fix is working
  • Tests multi-server distribution: Runs EC encoding with 6 volume servers to test shard distribution
  • Validates cleanup: Ensures original volumes are properly cleaned up after EC encoding

TestECEncodingMasterTimingRaceCondition

A focused test that specifically targets the master metadata timing race condition:

  • Simulates the exact race condition: Tests volume location collection timing relative to master metadata updates
  • Detects timing fix: Verifies that volume locations are collected BEFORE EC encoding starts
  • Demonstrates bug impact: Shows what happens when volume locations are unavailable after master metadata update

TestECEncodingRegressionPrevention

Regression tests that ensure:

  • Function signatures: Fixed functions still exist and return proper errors
  • Timing patterns: Volume location collection happens in the correct order

Test Architecture

The tests use:

  • Real SeaweedFS cluster: 1 master server + 6 volume servers
  • Multi-server setup: Tests realistic EC shard distribution across multiple servers
  • Timing simulation: Goroutines and delays to simulate race conditions
  • Output validation: Checks for specific log messages that prove the fix is working

Why Integration Tests Were Necessary

Unit tests could not catch this bug because:

  1. Race condition: The bug only occurred in real-world timing scenarios
  2. Master-volume server interaction: Required actual master metadata updates
  3. File system operations: Needed real volume creation and EC shard generation
  4. Cleanup timing: Required testing the sequence of operations in correct order

The integration tests successfully catch the timing bug by:

  • Testing real command execution: Uses actual ec.encode shell command
  • Simulating race conditions: Creates timing scenarios that expose the bug
  • Validating output messages: Checks for the key "Collecting volume locations...before EC encoding" message
  • Monitoring cleanup behavior: Ensures original volumes are properly deleted

Running the Tests

# Run all integration tests
go test -v

# Run only the main timing test
go test -v -run TestECEncodingVolumeLocationTimingBug

# Run only the race condition test
go test -v -run TestECEncodingMasterTimingRaceCondition

# Skip integration tests (short mode)
go test -v -short

Manual Testing with Makefile

A Makefile is provided for manual EC testing.

Requirements: curl, jq (command-line JSON processor)

# Quick start: start cluster and populate data
make setup

# Open weed shell to run EC commands
make shell

# Individual targets
make start      # Start test cluster (master + 6 volume servers + filer)
make stop       # Stop test cluster
make populate   # Populate ~300MB of test data
make status     # Show cluster and EC shard status
make clean      # Stop cluster and remove all test data
make help       # Show all targets

EC Rebalance Limited Slots (Unit Test)

The "no free ec shard slots" issue is tested with a unit test that works directly on topology data structures without requiring a running cluster.

Location: weed/shell/ec_rebalance_slots_test.go

Tests included:

  • TestECRebalanceWithLimitedSlots: Tests a topology with 6 servers, 7 EC volumes (98 shards)
  • TestECRebalanceZeroFreeSlots: Reproduces the exact 0 free slots scenario

Known Issue: When volume servers are at capacity (volumeCount == maxVolumeCount), the rebalance step fails with "no free ec shard slots" instead of recognizing that moving shards frees slots on source servers.

Test Results

With the fix: Shows "Collecting volume locations for N volumes before EC encoding..." message Without the fix: No collection message, potential timing race condition

The tests demonstrate that the fix prevents the volume location timing bug that caused double storage usage in EC encoding operations.