fix(tier): handle mutation intent CAS races (#5227)

* test(scripts): expand internode grpc ab env coverage

* fix(tier): handle mutation intent CAS races

Make tier mutation intent advancement retry idempotently when an If-Match CAS loses to a matching committed peer update.

Expand four-node inline fallback E2E coverage for mixed msgpack controls, transition readiness evidence, and per-node environment overrides.

Co-Authored-By: heihutu <heihutu@gmail.com>

---------

Co-authored-by: heihutu <heihutu@gmail.com>
This commit is contained in:
houseme
2026-07-25 19:10:13 +08:00
committed by GitHub
parent a63b79004c
commit 8e83087ba4
7 changed files with 747 additions and 96 deletions
+31
View File
@@ -15,11 +15,23 @@ trap cleanup EXIT
BEFORE_ENV="${TMP_DIR}/p2-before/server-env.sh"
AFTER_ENV="${TMP_DIR}/p2-after/server-env.sh"
P0_BEFORE_ENV="${TMP_DIR}/p0-before/server-env.sh"
P0_AFTER_ENV="${TMP_DIR}/p0-after/server-env.sh"
P1_BEFORE_ENV="${TMP_DIR}/p1-before/server-env.sh"
P1_AFTER_ENV="${TMP_DIR}/p1-after/server-env.sh"
P3_BEFORE_ENV="${TMP_DIR}/p3-before/server-env.sh"
P3_AFTER_ENV="${TMP_DIR}/p3-after/server-env.sh"
rg -qx 'export RUSTFS_INTERNODE_RPC_MSGPACK_ONLY=false' "$BEFORE_ENV"
rg -qx 'export RUSTFS_INTERNODE_RPC_MSGPACK_ONLY_FLEET_CONFIRMED=false' "$BEFORE_ENV"
rg -qx 'export RUSTFS_INTERNODE_RPC_MSGPACK_ONLY=true' "$AFTER_ENV"
rg -qx 'export RUSTFS_INTERNODE_RPC_MSGPACK_ONLY_FLEET_CONFIRMED=true' "$AFTER_ENV"
"$RUNNER" --stage p0 --phase before --out-root "$TMP_DIR" --dry-run >/dev/null
"$RUNNER" --stage p0 --phase after --out-root "$TMP_DIR" --dry-run >/dev/null
"$RUNNER" --stage p1 --phase before --out-root "$TMP_DIR" --dry-run >/dev/null
"$RUNNER" --stage p1 --phase after --out-root "$TMP_DIR" --dry-run >/dev/null
"$RUNNER" --stage p3 --phase before --out-root "$TMP_DIR" --dry-run >/dev/null
"$RUNNER" --stage p3 --phase after --out-root "$TMP_DIR" --dry-run >/dev/null
if grep -q 'RUSTFS_INTERNODE_RPC_MSGPACK_ONLY=true' "$BEFORE_ENV"; then
echo "p2 before must not request msgpack-only" >&2
@@ -31,4 +43,23 @@ if grep -q 'RUSTFS_INTERNODE_RPC_MSGPACK_ONLY_FLEET_CONFIRMED=true' "$BEFORE_ENV
exit 1
fi
rg -qx 'export RUSTFS_INTERNODE_RPC_TCP_NODELAY=false' "$P0_BEFORE_ENV"
rg -qx 'export RUSTFS_INTERNODE_RPC_HTTP2_STREAM_WINDOW_SIZE=0' "$P0_BEFORE_ENV"
rg -qx 'export RUSTFS_INTERNODE_RPC_HTTP2_CONN_WINDOW_SIZE=0' "$P0_BEFORE_ENV"
rg -qx 'export RUSTFS_INTERNODE_RPC_MAX_MESSAGE_SIZE=4194304' "$P0_BEFORE_ENV"
rg -qx '# \(cluster defaults — no overrides needed for this stage/phase\)' "$P0_AFTER_ENV"
rg -qx 'export RUSTFS_INTERNODE_CHANNEL_ISOLATION=false' "$P1_BEFORE_ENV"
rg -qx 'export RUSTFS_INTERNODE_CHANNEL_ISOLATION=true' "$P1_AFTER_ENV"
rg -qx 'export RUSTFS_INTERNODE_BULK_CHANNELS=2' "$P1_AFTER_ENV"
if rg -q 'export RUSTFS_INTERNODE_BULK_CHANNELS=' "$P1_BEFORE_ENV"; then
echo "p1 before should not pin an explicit bulk channel count" >&2
exit 1
fi
rg -qx 'export RUSTFS_INTERNODE_PREWARM=true' "$P3_AFTER_ENV"
rg -qx 'export RUSTFS_INTERNODE_OFFLINE_BYPASS=true' "$P3_AFTER_ENV"
rg -qx 'export RUSTFS_INTERNODE_OFFLINE_REPROBE_SECS=5' "$P3_AFTER_ENV"
rg -qx 'export RUSTFS_INTERNODE_OFFLINE_FAILURE_THRESHOLD=3' "$P3_AFTER_ENV"
echo "internode grpc A/B bench env tests passed"