From 4ec4d4dad0ebefe1d9ca5f2967e382ae062a541c Mon Sep 17 00:00:00 2001
From: Jim Garlick <garlick.jim@gmail.com>
Date: Tue, 4 Jul 2023 09:08:33 -0700
Subject: [PATCH] testsuite: add sched-simple alloc check test

Problem: there is no test script specifically for checking that
sched-simple does not double-book resources.

Add t2304-sched-simple-alloc-check.t which uses the
alloc-check.so jobtap plugin.

Currently this just validates the alloc-check plugin and checks
that sched-simple doesn't suffer from the same bug as
flux-framework/flux-sched#1043 but other tests could be added as
needed.
---
 t/Makefile.am                      |  1 +
 t/t2304-sched-simple-alloc-check.t | 79 ++++++++++++++++++++++++++++++
 2 files changed, 80 insertions(+)
 create mode 100755 t/t2304-sched-simple-alloc-check.t

diff --git a/t/Makefile.am b/t/Makefile.am
index ae216db81f1f..a0b273ab8a11 100644
--- a/t/Makefile.am
+++ b/t/Makefile.am
@@ -159,6 +159,7 @@ TESTSCRIPTS = \
 	t2300-sched-simple.t \
 	t2302-sched-simple-up-down.t \
 	t2303-sched-hello.t \
+	t2304-sched-simple-alloc-check.t \
 	t2310-resource-module.t \
 	t2311-resource-drain.t \
 	t2312-resource-exclude.t \
diff --git a/t/t2304-sched-simple-alloc-check.t b/t/t2304-sched-simple-alloc-check.t
new file mode 100755
index 000000000000..893d68182078
--- /dev/null
+++ b/t/t2304-sched-simple-alloc-check.t
@@ -0,0 +1,79 @@
+#!/bin/sh
+
+test_description='check that sched-simple never double books resources'
+
+# Append --logfile option if FLUX_TESTS_LOGFILE is set in environment:
+test -n "$FLUX_TESTS_LOGFILE" && set -- "$@" --logfile
+. $(dirname $0)/sharness.sh
+
+test_under_flux 1
+
+# Verify that alloc-check plugin works using alloc-bypass
+test_expect_success 'load alloc-bypass and alloc-check plugins' '
+	flux jobtap load alloc-bypass.so &&
+	flux jobtap load alloc-check.so
+'
+test_expect_success 'run an alloc-bypass sleep job' '
+	flux submit \
+	    -vvv \
+	    --wait-event=start \
+	    --setattr=alloc-bypass.R="$(flux R encode -r 0)" \
+	    -n 1 \
+	    sleep inf
+'
+test_expect_success 'a regular job fails with an alloc-check exception' '
+	run_timeout 30 flux submit --flags=waitable -vvv \
+	    --wait-event=exception \
+	    -N1 /bin/true >bypass.jobid
+'
+test_expect_success 'flux job wait says the job failed' '
+	test_must_fail flux job wait -v $(cat bypass.jobid)
+'
+test_expect_success 'clean up jobs' '
+	flux cancel --all &&
+	flux queue drain
+'
+test_expect_success 'unload plugins' '
+	flux jobtap remove alloc-check.so &&
+	flux jobtap remove alloc-bypass.so
+'
+
+# Check that sched-simple doesn't suffer from time limit issue like
+#   flux-framework/flux-sched#1043
+#
+test_expect_success 'configure epilog with 2s delay' '
+	flux config load <<-EOT
+	[job-manager]
+	plugins = [
+	   { load = "perilog.so" },
+	]
+	epilog.command = [ "flux", "perilog-run", "epilog", "-e", "sleep,2" ]
+	EOT
+'
+test_expect_success 'load alloc-check plugin' '
+	flux jobtap load alloc-check.so
+'
+test_expect_success 'submit consecutive jobs that exceed their time limit' '
+	(for i in $(seq 5); \
+	    do flux submit -N1 -x -t1s sleep inf; \
+	done) >jobids
+'
+test_expect_success 'wait for jobs to complete and capture their stderr' '
+	(for id in $(cat jobids); do \
+	    flux job attach $id || true; \
+	done) 2>joberr
+'
+test_expect_success 'some jobs received timeout exception' '
+	grep "job.exception type=timeout" joberr
+'
+test_expect_success 'no jobs received alloc-check exception' '
+	test_must_fail grep "job.exception type=alloc-check" joberr
+'
+test_expect_success 'remove alloc-check plugin' '
+	flux jobtap remove alloc-check.so
+'
+test_expect_success 'undo epilog config' '
+	flux config load </dev/null
+'
+
+test_done