Skip to content
Snippets Groups Projects
Commit aa5714d1 authored by Danny Auble's avatar Danny Auble
Browse files

better sacct test

parent 5c117cd1
No related branches found
No related tags found
No related merge requests found
......@@ -300,7 +300,7 @@ test12.# Testing of sacct command and options
===============================================
test12.1 Test sacct --help option.
test12.2 Test validity/accuracy of accounting data for exit code,
memory and real-time information.
memory and real-time information along with stating a running job.
(There are many more tests that should probably be added, but HP
is taking responsibility for validating this code, so we'll stick
with the basics here.)
......
#!/usr/bin/expect
############################################################################
# Purpose: Test of SLURM functionality
# Test basic sacct functionality and accuracy.
# Test sacct functionality and accuracy.
#
# Output: "TEST: #.#" followed by "SUCCESS" if test was successful, OR
# "FAILURE: ..." otherwise with an explanation of the failure, OR
......@@ -33,13 +33,16 @@ source ./globals
set test_id "12.2"
set exit_code 0
set file_in "test$test_id.input"
set file_out "test$test_id.output"
set file_err "test$test_id.error"
set file_prog "test$test_id.prog"
set job_id 0
set matches 0
# job paramters
set mem_size 100000
set sleep_time 31
set sleep_time 10
set ret_code 42
print_header $test_id
......@@ -47,15 +50,10 @@ print_header $test_id
#
# Check if accounting is enabled
#
set frequency 9999
set supported 0
log_user 0
spawn $scontrol show config
expect {
-re "JobAcctFrequency *= ($number)" {
set frequency $expect_out(1,string)
exp_continue
}
-re "JobAcctType *= jobacct/linux" {
set supported 1
exp_continue
......@@ -74,28 +72,33 @@ if {$supported == 0} {
exit 0
}
#
# Delete left-over program and rebuild it
#
exec $bin_rm -f $file_prog
exec $bin_rm -f $file_in $file_out $file_err
exec $bin_make -f /dev/null $file_prog
exec $bin_chmod 700 $file_prog
make_bash_script $file_in "
$srun $file_prog $ret_code $sleep_time $mem_size
"
#
# Run a simple job
# Usage: test12.2.prog <exit_code> <sleep_secs> <mem_kb>
#
set timeout [expr $max_job_delay + $sleep_time]
spawn $srun --output=none --error=none -v -t2 -N1 $file_prog $ret_code $sleep_time $mem_size
spawn $srun --batch --output=$file_out --error=$file_err -t2 $file_in
expect {
-re "launching ($number)\.0" {
-re "jobid ($number) submitted" {
set job_id $expect_out(1,string)
exp_continue
}
-re "Exited with exit code" {
send_user "This error was expected, no worries\n\n"
exp_continue
}
timeout {
send_user "\nFAILURE: srun not responding\n"
kill_srun
......@@ -106,13 +109,88 @@ expect {
}
}
if {$job_id == 0} {
send_user "\nFAILURE: job initiation failed\n"
send_user "\nFAILURE: batch submit failure\n"
exit 1
}
#
# Wait for job to run
#
if {[wait_for_job $job_id "RUNNING"] != 0} {
send_user "\nFAILURE: waiting for job to run\n"
set exit_code 1
}
# Wait for data to get logged
exec $bin_sleep 5
set mem_used -1
set mem_task -1
set ave_used -1
spawn $sacct --noheader --stat --job=$job_id.0 --fields vsize
expect {
-re "($float)(\[KM\]*)/($number) - ($float)(\[KM\]*)" {
set mem_used $expect_out(1,string)
set scale1 $expect_out(2,string)
set mem_task $expect_out(3,string)
set ave_used $expect_out(4,string)
set scale2 $expect_out(5,string)
if {[string compare $scale1 "M"] == 0} {
set mem_used [expr $mem_used * 1024]
} elseif {[string compare $scale1 "K"] == 0} {
set mem_used [expr $mem_used * 1]
} else {
set mem_used [expr $mem_used / 1024]
}
if {[string compare $scale2 "M"] == 0} {
set ave_used [expr $ave_used * 1024]
} elseif {[string compare $scale2 "K"] == 0} {
set ave_used [expr $ave_used * 1]
} else {
set ave_used [expr $ave_used / 1024]
}
exp_continue
}
timeout {
send_user "\nFAILURE: sacct not responding\n"
set exit_code 1
}
eof {
wait
}
}
if { $mem_used == -1 } {
send_user "\nFAILURE: sacct stat memory not found\n"
set exit_code 1
} elseif { $mem_task != 0 } {
send_user "\nFAILURE: sacct stat memory task not found\n"
set exit_code 1
} elseif { $ave_used != $mem_used } {
send_user "\nFAILURE: sacct stat memory task not equal to ave memory\n"
set exit_code 1
}
# Compute error in KB
set diff_mem [expr $mem_used - $mem_size]
set error_mem [expr abs($diff_mem)]
if {$error_mem > 4000} {
send_user "\nFAILURE: sacct memory use error of $error_mem KB\n"
send_user " Wanted $mem_size KB, got $mem_used KB\n"
set exit_code 1
} else {
send_user "\nSUCCESS: sacct memory use error of $error_mem KB\n"
}
#
# Wait for job to complete
#
if {[wait_for_job $job_id "DONE"] != 0} {
send_user "\nFAILURE: waiting for job to complete\n"
set exit_code 1
}
#
# Report basic sacct info
#
......@@ -232,9 +310,7 @@ if { $mem_used == -1 } {
# Compute error in KB
set diff_mem [expr $mem_used - $mem_size]
set error_mem [expr abs($diff_mem)]
if {$frequency >= $sleep_time} {
send_user "\nWARNING: JobAcctFrequency ($frequency) is too large for validating memory use\n"
} elseif {$error_mem > 4000} {
if {$error_mem > 4000} {
send_user "\nFAILURE: sacct memory use error of $error_mem KB\n"
send_user " Wanted $mem_size KB, got $mem_used KB\n"
set exit_code 1
......
0% Loading or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment