test(e2e): add distributed cluster regression coverage (#7158)

* test(e2e): add distributed 4x4 validation

* test(e2e): prove operations overlap data movement

---------

Co-authored-by: Zhengchao An <anzhengchao@gmail.com>
This commit is contained in:
RustFS
2026-09-06 10:12:06 +08:00
committed by GitHub
parent 6655272c90
commit 07833379b4
27 changed files with 4101 additions and 2 deletions
+2
View File
@@ -0,0 +1,2 @@
sha256-linux=9785867929047dfd8c6f768e0d2b1e0a8fdba85216f4a4139093b1619d03ff07
sha256-darwin=9785867929047dfd8c6f768e0d2b1e0a8fdba85216f4a4139093b1619d03ff07
+33
View File
@@ -183,6 +183,13 @@ test-group = 'e2e-reliability'
filter = 'package(e2e_test) & test(/^inline_fast_path_cluster_test::/)'
test-group = 'e2e-inline-boundaries'
# 4-node 4-drive distributed Actions suite: each case starts four rustfs
# processes and up to sixteen data directories. Serialize across nextest's
# process boundary so several 4x4 clusters never overlap.
[[profile.default.overrides]]
filter = 'package(e2e_test) & test(/^distributed::/)'
test-group = 'e2e-cluster-nightly'
# Vault KMS tests share the fixed dev-server port 8200. serial_test's #[serial]
# does not cross nextest process boundaries, so keep every Vault-backed test in
# one group.
@@ -526,6 +533,27 @@ path = "junit.xml"
filter = 'package(e2e_test)'
test-group = 'e2e-cluster-nightly'
# ---------------------------------------------------------------------------
# e2e-distributed profile — 4-node 4-disk Actions suite
# ---------------------------------------------------------------------------
# Storage-sensitive PR / nightly / dispatch lane owned by
# .github/workflows/e2e-distributed.yml.
# Each case starts four rustfs processes (and for site replication, two
# clusters). Upgrade cases also require RUSTFS_UPGRADE_SOURCE_BINARY.
# Serialized via e2e-cluster-nightly with no retries.
[profile.e2e-distributed]
default-filter = 'package(e2e_test) & test(/^distributed::/)'
fail-fast = false
# Decommission / rebalance cases poll for up to 180s with little stdout.
slow-timeout = { period = "120s", terminate-after = 6 }
[profile.e2e-distributed.junit]
path = "junit.xml"
[[profile.e2e-distributed.overrides]]
filter = 'package(e2e_test)'
test-group = 'e2e-cluster-nightly'
# ---------------------------------------------------------------------------
# e2e-odm-interop profile — on-demand migration provider interop lane (ODM-20)
# ---------------------------------------------------------------------------
@@ -586,6 +614,10 @@ path = "junit.xml"
# cluster-fault lane. heal_erasure_disk_rebuild is intentionally not
# excluded here because backlog#2213 promotes core heal rebuild coverage to
# this merge/main lane while retaining nightly coverage.
# * distributed:: — 4-node 4-disk Actions suite (S3, lock, versioning,
# replication, quota, observability, expand/decommission/rebalance, site
# replication, chaos, upgrade history/IAM). Owns [profile.e2e-distributed] and
# .github/workflows/e2e-distributed.yml.
# * on_demand_migration::interop_test — the ODM-20 provider interoperability
# cases, which are meaningless without a source: they run in the dedicated
# [profile.e2e-odm-interop] lane below, where the workflow points them at a
@@ -607,6 +639,7 @@ default-filter = """
package(e2e_test)
& !test(/^protocols::/)
& !test(/^(admin_timeout_regression_test|cluster_concurrency_test|cluster_multidrive_pool_test|degraded_listing_availability_test|namespace_lock_quorum_test|object_lambda_test|stale_multipart_cleanup_cluster_test)::/)
& !test(/^distributed::/)
& !test(/^replication_extension_test::/)
& !test(/^replication_target_matrix_test::/)
& !test(/^on_demand_migration::(concurrency_test|fault_test|interop_test|real_source_test)::/)