Split miri into two parts, run slow only in nightly

Author: def-

ci/nightly/pipeline.yml

@@ -17,6 +17,7 @@ steps:
       - select: Tests
         key: tests
         options:
+          - { value: miri-test }
           - { value: kafka-matrix }
           - { value: kafka-multi-broker }
           - { value: redpanda-testdrive }
@@ -92,6 +93,18 @@ steps:
 
   - wait: ~
 
+  - id: miri-test
+    label: Miri test (full)
+    timeout_in_minutes: 600
+    artifact_paths: [junit_*.xml, target/nextest/ci/junit_cargo-test.xml]
+    plugins:
+      - ./ci/plugins/scratch-aws-access: ~
+      - ./ci/plugins/mzcompose:
+          composition: cargo-test
+          args: [--miri-full]
+    agents:
+      queue: builder-linux-x86_64
+
   - id: feature-benchmark
     label: "Feature benchmark against 'latest'"
     timeout_in_minutes: 360

ci/test/cargo-test-miri-fast.sh

@@ -0,0 +1,26 @@
+#!/usr/bin/env bash
+
+# Copyright Materialize, Inc. and contributors. All rights reserved.
+#
+# Use of this software is governed by the Business Source License
+# included in the LICENSE file at the root of this repository.
+#
+# As of the Change Date specified in that file, in accordance with
+# the Business Source License, use of this software will be governed
+# by the Apache License, Version 2.0.
+#
+# cargo-test-miri.sh — runs subset of unit tests under miri to check for
+# undefined behaviour.
+
+set -euo pipefail
+
+# miri artifacts are thoroughly incompatible with normal build artifacts,
+# so keep them away from the `target` directory.
+export CARGO_TARGET_DIR="$PWD/miri-target"
+export MIRIFLAGS="-Zmiri-disable-isolation -Zmiri-strict-provenance"
+
+PARTITION=$((${BUILDKITE_PARALLEL_JOB:-0}+1))
+TOTAL=${BUILDKITE_PARALLEL_JOB_COUNT:-1}
+
+# exclude network-based and more complex tests which run out of memory
+cargo miri nextest run -j"$(nproc)" --partition "count:$PARTITION/$TOTAL" --no-fail-fast --workspace --exclude 'mz-adapter*' --exclude 'mz-environmentd*' --exclude 'mz-expr*' --exclude 'mz-compute-client*' --exclude 'mz-persist-client*' --exclude 'mz-ssh-util*' --exclude 'mz-rocksdb*' --exclude 'mz-sqllogictest*'

ci/test/cargo-test/mzcompose.py

@@ -38,7 +38,8 @@
 
 
 def workflow_default(c: Composition, parser: WorkflowArgumentParser) -> None:
-    parser.add_argument("--miri", action="store_true")
+    parser.add_argument("--miri-full", action="store_true")
+    parser.add_argument("--miri-fast", action="store_true")
     parser.add_argument("args", nargs="*")
     args = parser.parse_args()
     c.up("zookeeper", "kafka", "schema-registry", "postgres", "cockroach")
@@ -108,11 +109,16 @@ def workflow_default(c: Composition, parser: WorkflowArgumentParser) -> None:
                 ["buildkite-agent", "artifact", "upload", "coverage/cargotest.lcov.xz"]
             )
     else:
-        if args.miri:
+        if args.miri_full:
             spawn.runv(
                 ["bin/ci-builder", "run", "nightly", "ci/test/cargo-test-miri.sh"],
                 env=env,
             )
+        elif args.miri_fast:
+            spawn.runv(
+                ["bin/ci-builder", "run", "nightly", "ci/test/cargo-test-miri-fast.sh"],
+                env=env,
+            )
         else:
             spawn.runv(
                 [

ci/test/pipeline.template.yml

@@ -155,17 +155,16 @@ steps:
       queue: builder-linux-x86_64
 
   - id: miri-test
-    label: Miri test %n
-    command: bin/ci-builder run nightly ci/test/cargo-test-miri.sh
+    label: Miri test (fast) %n
     inputs: [src]
-    parallelism: 6
+    parallelism: 2
     timeout_in_minutes: 30
     artifact_paths: [junit_*.xml, target/nextest/ci/junit_cargo-test.xml]
     plugins:
       - ./ci/plugins/scratch-aws-access: ~
       - ./ci/plugins/mzcompose:
           composition: cargo-test
-          args: [--miri]
+          args: [--miri-fast]
     agents:
       queue: builder-linux-x86_64
     coverage: skip