Skip to content
GitLab
Explore
Sign in
Primary navigation
Search or go to…
Project
S
Slurm
Manage
Activity
Members
Labels
Plan
Issues
Issue boards
Milestones
Code
Merge requests
Repository
Branches
Commits
Tags
Repository graph
Compare revisions
Snippets
Build
Pipelines
Jobs
Pipeline schedules
Artifacts
Deploy
Releases
Package Registry
Model registry
Operate
Environments
Terraform modules
Monitor
Incidents
Service Desk
Analyze
Value stream analytics
Contributor analytics
CI/CD analytics
Repository analytics
Model experiments
Help
Help
Support
GitLab documentation
Compare GitLab plans
Community forum
Contribute to GitLab
Provide feedback
Terms and privacy
Keyboard shortcuts
?
Snippets
Groups
Projects
Show more breadcrumbs
tud-zih-energy
Slurm
Commits
8c4fdffe
Commit
8c4fdffe
authored
5 years ago
by
Brian Christiansen
Browse files
Options
Downloads
Patches
Plain Diff
Update test7.20 to catch passing/failing het jobs
when one offset passes and other fails. Bug 6892
parent
1460a6b5
No related branches found
No related tags found
No related merge requests found
Changes
2
Hide whitespace changes
Inline
Side-by-side
Showing
2 changed files
testsuite/expect/test7.20
+50
-33
50 additions, 33 deletions
testsuite/expect/test7.20
testsuite/expect/test7.20_scripts/reject.lua
+3
-0
3 additions, 0 deletions
testsuite/expect/test7.20_scripts/reject.lua
with
53 additions
and
33 deletions
testsuite/expect/test7.20
+
50
−
33
View file @
8c4fdffe
...
@@ -176,7 +176,7 @@ proc check_rc { exit_code } {
...
@@ -176,7 +176,7 @@ proc check_rc { exit_code } {
}
}
}
}
proc test_bin_pass { bin args count } {
proc test_bin_pass { bin args
het_job
count } {
global salloc ta1 tq2 srun number
global salloc ta1 tq2 srun number
set got_init 0
set got_init 0
...
@@ -193,15 +193,21 @@ proc test_bin_pass { bin args count } {
...
@@ -193,15 +193,21 @@ proc test_bin_pass { bin args count } {
}
}
# salloc: 0: submit1
# salloc: 0: submit1
-re "^(srun|salloc|sbatch): ($number: |)submit1\r\n" {
-re "^(srun|salloc|sbatch): ($number: |)submit1\r\n" {
incr got_sub1
if {!$het_job || $expect_out(2,string) != ""} {
incr got_sub1
}
exp_continue
exp_continue
}
}
-re "^(srun|salloc|sbatch): ($number: |)submit2\r\n" {
-re "^(srun|salloc|sbatch): ($number: |)submit2\r\n" {
incr got_sub2
if {!$het_job || $expect_out(2,string) != ""} {
incr got_sub2
}
exp_continue
exp_continue
}
}
-re "^(srun|salloc|sbatch): ($number: |)submit3\r\n" {
-re "^(srun|salloc|sbatch): ($number: |)submit3\r\n" {
incr got_sub3
if {!$het_job || $expect_out(2,string) != ""} {
incr got_sub3
}
exp_continue
exp_continue
}
}
timeout {
timeout {
...
@@ -219,14 +225,16 @@ proc test_bin_pass { bin args count } {
...
@@ -219,14 +225,16 @@ proc test_bin_pass { bin args count } {
log_info "$bin rc:$rc init:$got_init sub1:$got_sub1 sub2:$got_sub2 sub3:$got_sub3 expected:$count"
log_info "$bin rc:$rc init:$got_init sub1:$got_sub1 sub2:$got_sub2 sub3:$got_sub3 expected:$count"
if {$rc != 0 || $got_sub1 != $count || $got_sub2 != $count ||
if {($rc != 0) ||
$got_sub3 != $count} {
($got_sub1 != $count) ||
($got_sub2 != $count) ||
($got_sub3 != $count)} {
log_error "Invalid $bin response"
log_error "Invalid $bin response"
endit 1
endit 1
}
}
}
}
proc test_bin_fail { bin args count } {
proc test_bin_fail { bin args
het_job
count } {
global ta1 tq1 tq2 srun number
global ta1 tq1 tq2 srun number
set got_init 0
set got_init 0
...
@@ -244,15 +252,21 @@ proc test_bin_fail { bin args count } {
...
@@ -244,15 +252,21 @@ proc test_bin_fail { bin args count } {
#srun: error: submit1\r\n
#srun: error: submit1\r\n
#salloc: error: 0: submit1\r\n
#salloc: error: 0: submit1\r\n
-re "^(srun|salloc|sbatch): error: ($number: |)submit1\r\n" {
-re "^(srun|salloc|sbatch): error: ($number: |)submit1\r\n" {
incr got_sub1
if {!$het_job || $expect_out(2,string) != ""} {
incr got_sub1
}
exp_continue
exp_continue
}
}
-re "^(srun|salloc|sbatch): error: ($number: |)submit2\r\n" {
-re "^(srun|salloc|sbatch): error: ($number: |)submit2\r\n" {
incr got_sub2
if {!$het_job || $expect_out(2,string) != ""} {
incr got_sub2
}
exp_continue
exp_continue
}
}
-re "^(srun|salloc|sbatch): error: ($number: |)submit3\r\n" {
-re "^(srun|salloc|sbatch): error: ($number: |)submit3\r\n" {
incr got_sub3
if {!$het_job || $expect_out(2,string) != ""} {
incr got_sub3
}
exp_continue
exp_continue
}
}
timeout {
timeout {
...
@@ -270,8 +284,10 @@ proc test_bin_fail { bin args count } {
...
@@ -270,8 +284,10 @@ proc test_bin_fail { bin args count } {
log_info "$bin rc:$rc init:$got_init sub1:$got_sub1 sub2:$got_sub2 sub3:$got_sub3 expected:$count"
log_info "$bin rc:$rc init:$got_init sub1:$got_sub1 sub2:$got_sub2 sub3:$got_sub3 expected:$count"
if {$rc == 0 || $got_sub1 != $count || $got_sub2 != $count ||
if {($rc == 0) ||
$got_sub3 != $count} {
($got_sub1 != $count) ||
($got_sub2 != $count) ||
($got_sub3 != $count)} {
log_error "Invalid $bin response"
log_error "Invalid $bin response"
endit 1
endit 1
}
}
...
@@ -328,13 +344,14 @@ sleep 1
...
@@ -328,13 +344,14 @@ sleep 1
file mtime $config_dir/job_submit.lua [timestamp]
file mtime $config_dir/job_submit.lua [timestamp]
# Check that all job types are rejected
# Check that all job types are rejected
test_bin_fail $salloc {-A $ta1 --qos $tq2 -n5 "/bin/true"} 1
test_bin_fail $salloc {-A $ta1 --qos $tq2 -n5 "/bin/true"} 0 1
test_bin_fail $salloc {-A $ta1 --qos $tq2 -n5 : -n3 : -n1 "/bin/true"} 1
test_bin_fail $salloc {-A $ta1 --qos $tq2 -n5 : -n3 : -n1 "/bin/true"} 1 1
test_bin_fail $srun {-A $ta1 --qos $tq2 -n5 /bin/true} 1
test_bin_fail $srun {-A $ta1 --qos $tq2 -n5 /bin/true} 0 1
test_bin_fail $srun {-A $ta1 --qos $tq2 -n5 : -n3 : -n1 /bin/true} 1
test_bin_fail $srun {-A $ta1 --qos $tq2 -n5 : -n3 : -n1 /bin/true} 1 1
test_bin_fail $sbatch {-W -A $ta1 --qos $tq2 -n5 --wrap /bin/true} 1
test_bin_fail $sbatch {-W -A $ta1 --qos $tq2 -n5 --wrap /bin/true} 0 1
test_bin_fail $sbatch {-W -A $ta1 --qos $tq2 -n5 --array 10 --wrap /bin/true} 1
test_bin_fail $sbatch {-W -A $ta1 --qos $tq2 -n5 --array 10 --wrap /bin/true} 0 1
test_bin_fail $sbatch {-W -A $ta1 --qos $tq2 -n5 : -n3 : -n1 --wrap /bin/true} 1
test_bin_fail $sbatch {-W -A $ta1 --qos $tq2 -n5 : -n3 : -n1 --wrap /bin/true} 1 1
test_bin_fail $sbatch {-W -A $ta1 --qos $tq2 -n1 -J pass : -n3 -J error : -n5 --wrap /bin/true} 1 2
do_copy $test_lua_pass $config_dir/job_submit.lua
do_copy $test_lua_pass $config_dir/job_submit.lua
# Sleep for 1 second to make sure that modify time is different from last copy.
# Sleep for 1 second to make sure that modify time is different from last copy.
...
@@ -342,22 +359,22 @@ sleep 1
...
@@ -342,22 +359,22 @@ sleep 1
file mtime $config_dir/job_submit.lua [timestamp]
file mtime $config_dir/job_submit.lua [timestamp]
# Check that passing works
# Check that passing works
test_bin_pass $salloc {-A $ta1 --qos $tq2 -n5 "/bin/true"} 1
test_bin_pass $salloc {-A $ta1 --qos $tq2 -n5 "/bin/true"}
0
1
test_bin_pass $salloc {-A $ta1 --qos $tq2 -n5 : -n3 : -n1 "/bin/true"} 3
test_bin_pass $salloc {-A $ta1 --qos $tq2 -n5 : -n3 : -n1 "/bin/true"}
1
3
test_bin_pass $srun {-A $ta1 --qos $tq2 -n5 /bin/true} 1
test_bin_pass $srun {-A $ta1 --qos $tq2 -n5 /bin/true}
0
1
test_bin_pass $srun {-A $ta1 --qos $tq2 -n5 : -n3 : -n1 /bin/true} 3
test_bin_pass $srun {-A $ta1 --qos $tq2 -n5 : -n3 : -n1 /bin/true}
1
3
test_bin_pass $sbatch {-W -A $ta1 --qos $tq2 -n5 --wrap /bin/true} 1
test_bin_pass $sbatch {-W -A $ta1 --qos $tq2 -n5 --wrap /bin/true}
0
1
test_bin_pass $sbatch {-W -A $ta1 --qos $tq2 -n5 --array 10 --wrap /bin/true} 1
test_bin_pass $sbatch {-W -A $ta1 --qos $tq2 -n5 --array 10 --wrap /bin/true}
0
1
test_bin_pass $sbatch {-W -A $ta1 --qos $tq2 -n5 : -n3 : -n1 --wrap /bin/true} 3
test_bin_pass $sbatch {-W -A $ta1 --qos $tq2 -n5 : -n3 : -n1 --wrap /bin/true}
1
3
# Check that messages are still sent with failing QOS but passing filter
# Check that messages are still sent with failing QOS but passing filter
test_bin_fail $salloc {-A $ta1 --qos $tq1 -n5 "/bin/true"} 1
test_bin_fail $salloc {-A $ta1 --qos $tq1 -n5 "/bin/true"}
0
1
test_bin_fail $salloc {-A $ta1 --qos $tq1 -n5 : -n3 : -n1 "/bin/true"} 1
test_bin_fail $salloc {-A $ta1 --qos $tq1 -n5 : -n3 : -n1 "/bin/true"} 1
1
test_bin_fail $srun {-A $ta1 --qos $tq1 -n5 /bin/true} 1
test_bin_fail $srun {-A $ta1 --qos $tq1 -n5 /bin/true}
0
1
test_bin_fail $srun {-A $ta1 --qos $tq1 -n5 : -n3 : -n1 /bin/true} 1
test_bin_fail $srun {-A $ta1 --qos $tq1 -n5 : -n3 : -n1 /bin/true} 1
1
test_bin_fail $sbatch {-W -A $ta1 --qos $tq1 -n5 --wrap /bin/true} 1
test_bin_fail $sbatch {-W -A $ta1 --qos $tq1 -n5 --wrap /bin/true}
0
1
test_bin_fail $sbatch {-W -A $ta1 --qos $tq1 -n5 --array 10 --wrap /bin/true} 1
test_bin_fail $sbatch {-W -A $ta1 --qos $tq1 -n5 --array 10 --wrap /bin/true}
0
1
test_bin_fail $sbatch {-W -A $ta1 --qos $tq1 -n5 : -n3 : -n1 --wrap /bin/true} 3
test_bin_fail $sbatch {-W -A $ta1 --qos $tq1 -n5 : -n3 : -n1 --wrap /bin/true}
1
3
check_rc $exit_code
check_rc $exit_code
...
...
This diff is collapsed.
Click to expand it.
testsuite/expect/test7.20_scripts/reject.lua
+
3
−
0
View file @
8c4fdffe
...
@@ -6,6 +6,9 @@ function slurm_job_submit(job_desc, part_list, submit_uid)
...
@@ -6,6 +6,9 @@ function slurm_job_submit(job_desc, part_list, submit_uid)
slurm
.
log_user
(
"submit1"
)
slurm
.
log_user
(
"submit1"
)
slurm
.
log_user
(
"submit2"
)
slurm
.
log_user
(
"submit2"
)
slurm
.
log_user
(
"submit3"
)
slurm
.
log_user
(
"submit3"
)
if
(
job_desc
.
name
==
"pass"
)
then
return
slurm
.
SUCCESS
end
return
slurm
.
ERROR
return
slurm
.
ERROR
end
end
...
...
This diff is collapsed.
Click to expand it.
Preview
0%
Loading
Try again
or
attach a new file
.
Cancel
You are about to add
0
people
to the discussion. Proceed with caution.
Finish editing this message first!
Save comment
Cancel
Please
register
or
sign in
to comment