Skip to content

Commit

Permalink
Merge pull request #426 from perftool-incubator/cdmv7
Browse files Browse the repository at this point in the history 
Support Common Data Model "v7dev"
  • Loading branch information
@atheurer
atheurer authored Nov 6, 2023
2 parents 5e3ee9d + fcd4566 commit 3125528
Show file tree
Hide file tree
Showing 6 changed files with 155 additions and 13 deletions.
5 changes: 4 additions & 1 deletion endpoints/base
View file
Original file line number Diff line number Diff line change
Expand Up @@ -989,11 +989,14 @@ function process_roadblocks() {
roadblock_exit_on_error ${roadblock_rc}



do_roadblock "engine-init-begin" ${engine_script_start_timeout}
roadblock_rc=$?
roadblock_exit_on_error ${roadblock_rc}

do_roadblock "engine-init-end" ${engine_script_start_timeout}
call_endpoint_specific_function "${endpoint_type}" "engine_init" "engine-init-end"

do_roadblock "engine-init-end" ${engine_script_start_timeout} messages "${msg_file}"
roadblock_rc=$?
roadblock_exit_on_error ${roadblock_rc}

Expand Down
28 changes: 27 additions & 1 deletion endpoints/k8s/k8s
View file
Original file line number Diff line number Diff line change
Expand Up @@ -54,6 +54,7 @@ endpoint_name="k8s"
# TODO: instead of using a prefix in the pods' names, use a unique k8s project
pod_prefix="rickshaw"
project_name="crucible-rickshaw"
hypervisor_host="none"
unique_project="0"
hostNetwork="0"
hugepage="0"
Expand Down Expand Up @@ -84,6 +85,31 @@ function cleanup_json() {
fi
}

function endpoint_k8s_engine_init() {
echo "Running endpoint_engine_init"
msg_file="$endpoint_run_dir/env-vars.json"
echo '[' >$msg_file
local count=0
for this_cs_label in $all_pods; do

hosted_by=`jq -r .spec.nodeName $endpoint_run_dir/kubectl-get-pod-$this_cs_label.json`

if [ $count -gt 0 ]; then
printf "," >>$msg_file
fi
echo '{"recipient":{"type":"follower","id":"'$this_cs_label'"},"user-object":{"env-vars":{' >>$msg_file
echo '"endpoint_label": "'$endpoint_label'",' >>$msg_file
echo '"hosted_by": "'$hosted_by'",' >>$msg_file
echo '"hypervisor_host": "'$hypervisor_host'",' >>$msg_file
echo '"userenv": "'$userenv'",' >>$msg_file
echo '"osruntime": "'$os_runtime'"' >>$msg_file
echo '}}}' >>$msg_file

let count=$count+1
done
echo ']' >>$msg_file
}

function endpoint_k8s_test_stop() {
local msgs_dir="$1"; shift
local test_id="$1"; shift
Expand Down Expand Up @@ -1233,7 +1259,7 @@ if [ "${masters_tool_collect}" == "1" ]; then
verify_pods_running tool_master_nodes $master_tool_pods
echo "These nodes are hosting the master-tool pods: $master_nodes"
new_k8s_followers+=" ${master_tool_pods}"
all_pods+=" $worker_tool_pods"
all_pods+=" $master_tool_pods"
fi

process_roadblocks k8s ${new_k8s_followers}
29 changes: 27 additions & 2 deletions endpoints/remotehost/remotehost
View file
Original file line number Diff line number Diff line change
Expand Up @@ -46,6 +46,30 @@ endpoint_name="remotehost"
image_cache_size=3
osruntime[default]="chroot"
host_mounts=""
hypervisor_host="none" # Default is no hypervisor

function endpoint_remotehost_engine_init() {
echo "Running endpoint_engine_init"

msg_file="$endpoint_run_dir/env-vars.json"
echo '[' >$msg_file
local count=0
for this_cs_label in ${clients[@]} ${servers[@]} ${collectors[@]}; do
if [ $count -gt 0 ]; then
printf "," >>$msg_file
fi
echo '{"recipient":{"type":"follower","id":"'$this_cs_label'"},"user-object":{"env-vars":{' >>$msg_file
echo '"endpoint_label": "'$endpoint_label'",' >>$msg_file
echo '"hosted_by": "'$host'",' >>$msg_file
echo '"hypervisor_host": "'$hypervisor_host'",' >>$msg_file
echo '"userenv": "'$userenv'",' >>$msg_file
echo '"osruntime": "'$os_runtime'"' >>$msg_file
echo '}}}' >>$msg_file

let count=$count+1
done
echo ']' >>$msg_file
}

function endpoint_remotehost_test_stop() {
echo "Running endpoint_remotehost_test_stop"
Expand Down Expand Up @@ -209,6 +233,9 @@ function process_remotehost_opts() {
controller_ipaddr=`get_controller_ip $host`
fi
;;
hypervisor-host)
hypervisor_host=$val
;;
controller-ip)
controller_ipaddr=$val
;;
Expand Down Expand Up @@ -354,7 +381,6 @@ function launch_osruntime() {
base_cmd+=" --cs-label=$this_cs_label"
base_cmd+=" --base-run-dir=$base_run_dir"
base_cmd+=" --endpoint=remotehost"
base_cmd+=" --osruntime=${os_runtime}"
base_cmd+=" --max-sample-failures=$max_sample_failures"
base_cmd+=" --max-rb-attempts=$max_rb_attempts"
base_cmd+=" --cpu-partitions=${total_cpu_partitions}"
Expand Down Expand Up @@ -383,7 +409,6 @@ function launch_osruntime() {
echo "cpu_partitions=${total_cpu_partitions}" >> ${endpoint_run_dir}/${env_file}
echo "cpu_partition_index=${count}" >> ${endpoint_run_dir}/${env_file}
echo "endpoint=remotehost" >> ${endpoint_run_dir}/${env_file}
echo "osruntime=${os_runtime}" >> ${endpoint_run_dir}/${env_file}
echo "max_sample_failures=$max_sample_failures" >> ${endpoint_run_dir}/${env_file}
echo "max_rb_attempts=$max_rb_attempts" >> ${endpoint_run_dir}/${env_file}
echo "ssh_id=${ssh_id}" >> ${endpoint_run_dir}/${env_file}
Expand Down
20 changes: 20 additions & 0 deletions engine/engine-script
View file
Original file line number Diff line number Diff line change
Expand Up @@ -65,6 +65,26 @@ roadblock_exit_on_error ${roadblock_rc}
do_roadblock engine-init-end ${engine_script_start_timeout}
roadblock_rc=$?
roadblock_exit_on_error ${roadblock_rc}
msgs_log_file="${roadblock_msgs_dir}/engine-init-end.json"
if [ -f ${msgs_log_file} ]; then
env_vars=`jq -r '.received[] | .payload.message."user-object"."env-vars" ' ${msgs_log_file} | grep -v null`
if [ ! -z "${env_vars}" ]; then
printf "Found new env vars:"
printf "%s\n" "${env_vars}"
eval `echo $env_vars | jq -r 'to_entries | .[] | "export " + .key + "=" + .value'`
echo $env_vars | jq -r 'to_entries | .[] | "export " + .key + "=" + .value'
else
printf "Could not find new env vars\n"
fi
else
printf "Could not find %s\n" ${msgs_log_file}
echo '#/bin/ls -l '$roadblock_msgs_dir
/bin/ls -l $roadblock_msgs_dir
fi

# There are a number of env vars needed for post-processing metrics
# Do this after engine-init, as some vars can be passed to the engine in that RB
env >engine-env.txt


do_roadblock get-data-begin ${default_timeout}
Expand Down
5 changes: 4 additions & 1 deletion engine/engine-script-library
View file
Original file line number Diff line number Diff line change
Expand Up @@ -1082,6 +1082,8 @@ function stop_tools() {
tools_disabled=${1}; shift

echo "running stop_tools()"
echo "pwd: `/bin/pwd`"
local old_dir=`/bin/pwd`

if [ "${tools_disabled}" == "1" ]; then
echo "Not stopping tools because --disable-tools=1 was used"
Expand All @@ -1097,7 +1099,8 @@ function stop_tools() {
eval $tool_cmd
tool_cmd_rc=$?
echo "Stopping tool '${tool_name}' with command '${tool_cmd}' returned ${tool_cmd_rc}"

echo "Copying $old_dir/engine-env.txt to `/bin/pwd`"
/bin/cp $old_dir/engine-env.txt .
popd >/dev/null
else
abort_error "stop_tools: Failed to pushd to ${tool_name}" stop-tools-end
Expand Down
81 changes: 73 additions & 8 deletions rickshaw-index
View file
Original file line number Diff line number Diff line change
Expand Up @@ -39,7 +39,7 @@ $toolbox::logging::debug = 0;

my @pids;
my $index_tools = 1;
my @suported_cdm_vers = ('v5dev', 'v6dev');
my @suported_cdm_vers = ('v6dev', 'v7dev');
my %result;
my $base_run_dir;
my %cdm = ( 'ver' => '' );
Expand Down Expand Up @@ -352,8 +352,9 @@ sub wait_for_metric_descs {
# with create_es_doc(). Metrics can be indexed from either a benchmark sample directory or a tool
# directory.
sub index_metrics {
my $index_or_queue = shift;
my $metr_file = shift; # filename without .json or .csv
my $index_or_queue = shift; # what action to take, index = submit to ES, queue = enqueue to file for bulk index later
my $metr_dir = shift; # directory where metric files exist
my $metr_file = shift; # metric filename without .json or .csv
my $cstype = shift;
my $csid = shift;
my $base_doc_ref = shift; # metric_desc doc gets populated with this, usually a run doc or period doc
Expand All @@ -364,6 +365,60 @@ sub index_metrics {
my $earliest_begin;
my $latest_end;
my $coder = JSON::XS->new->canonical;

my $dir = pushd($metr_dir);
my %eng_env_vars;
if ($cdm{'ver'} eq 'v7dev') {
# engine-type and engine-id replace cstype and csid, but cstype/id to be removed later
$eng_env_vars{'engine-type'} = $cstype;
$eng_env_vars{'engine-id'} = $csid;
# engine-role is the engine's primary purpose (to run a bvenchmark, to collect tool data, etc)
if ($cstype =~ /^client$|^server$/) {
$eng_env_vars{'engine-role'} = 'benchmarker';
$eng_env_vars{'benchmark-role'} = $cstype;
# the following to be properly defined in a future enhancement
$eng_env_vars{'benchmark-name'} = 'unknown';
$eng_env_vars{'benchmark-group'} = 'unknown';
# to be 'none' once one-tool-per engine is implemented (and a benchmark engine no longer runs tools)
$eng_env_vars{'tool-name'} = 'unknown';
} elsif ($cstype =~ /^worker$|^master$|^profiler$/) {
$eng_env_vars{'engine-role'} = 'profiler';
$eng_env_vars{'benchmark-role'} = 'none';
$eng_env_vars{'benchmark-name'} = 'none';
# the following to be properly defined once one-tool-per engine is implemented
$eng_env_vars{'tool-name'} = 'unknown';
}

# Note that CDM being v7 should not assume all metrics have engine_env.txt because one may be re-postprocessing older crucible runs
my $eng_env_file = "engine-env.txt";
if (! -e $eng_env_file) {
$eng_env_file .= ".xz";
}
if (! -e $eng_env_file) {
printf "pwd: %s\nCould not find %s, will not use engine env vars for metadata\n", getcwd(), $eng_env_file;
} else {
my $eng_env_fh = new IO::Uncompress::UnXz $eng_env_file, Transparent => 1 || die "[ERROR]could not open file " . $eng_env_file;
# Cull out env vars which we want as metadata
#printf "Looking for engine env vars\n";
my @varnames = ('HOSTNAME', 'engine_type', 'engine_role', 'benchmark_group', 'benchmark_role', 'hosted_by', 'hypervisor_host', 'osruntime', 'endpoint_label', 'userenv');
while (<$eng_env_fh>) {
chomp;
foreach my $varname (@varnames) {
if (/^$varname=(.*)$/) {
my $val = $1;
$varname =~ s/_/-/g;
if ($varname eq "HOSTNAME") {
$varname = "hostname";
}
$eng_env_vars{$varname} = $val;
next;
}
}
}
close $eng_env_fh;
}
}

# Copy data from 'parent' doc so querying directly for metric_desc with
# run data is possible
my $metr_json_file = $metr_file . ".json";
Expand All @@ -372,7 +427,7 @@ sub index_metrics {
$metr_json_file .= ".xz";
}
if (! -e $metr_json_file) {
printf "Could not find %s, exiting\n", $metr_json_file;
printf "pwd: %s\nCould not find %s, exiting\n", getcwd(), $metr_json_file;
exit 1;
}
if (! -e $metr_csv_file) {
Expand Down Expand Up @@ -410,11 +465,17 @@ sub index_metrics {
if ( exists $$this_metr{'values'} ) {
$metr_desc_doc{'metric_desc'}{'values'} = $$this_metr{'values'};
}
# this is where we add engine-related metadata
$metr_desc_doc{'metric_desc'}{'names'}{'cstype'} = $cstype;
$metr_desc_doc{'metric_desc'}{'names'}{'csid'} = $csid;
foreach my $env_var (keys %eng_env_vars) {
$metr_desc_doc{'metric_desc'}{'names'}{$env_var} = $eng_env_vars{$env_var};
}

my @names_list = sort(keys(%{ $metr_desc_doc{'metric_desc'}{'names'} }));
$metr_desc_doc{'metric_desc'}{'names-list'} = \@names_list;
my $metr_desc_doc_json = $coder->encode(\%metr_desc_doc);
#printf "metric_desc_doc:\n %s\n", $metr_desc_doc_json;
# We do not use index_es_doc() here because that requires getting all info from the %result,
# and %result (rickshaw-run.json) by design does not include any metric data, as it would be
# way too large.
Expand Down Expand Up @@ -619,6 +680,7 @@ if (ref $idx_resp_ref eq ref [] ) {
push(@all_indices, $idx_name);
}
my @sorted_vers = sort(keys(%vers));
printf "sorted_vers: @sorted_vers \n";
$latest_ver = $sorted_vers[0];
if (defined $latest_ver) {
printf "Latest CDM version found in local ES instance is %s\n", $latest_ver;
Expand Down Expand Up @@ -650,6 +712,7 @@ if (ref $idx_resp_ref eq ref [] ) {
}

$cdm{'ver'} = $latest_ver;
printf "Latest CDM version found in local ES instance is %s\n", $latest_ver;
if (not grep(/^$cdm{'ver'}$/, @suported_cdm_vers)) {
printf "The version of CDM used in ES [%s] is not one that is supported by rickshaw-index: [%s]\n",
$cdm{'ver'}, join(" ", @suported_cdm_vers);
Expand Down Expand Up @@ -702,7 +765,8 @@ if (-e $tool_dir and $index_tools == 1) {
$tool_file =~ s/(metric-data-\S+)\.json.*/$1/;
printf "Working on tool_file: %s\n", $tool_file;

my %job_args = ( 'tool-file' => $tool_dir . "/" . $tool_file,
my %job_args = ( 'tool-dir' => $tool_dir,
'tool-file' => $tool_file,
'collector' => $collector,
'num' => $num,
'doc-ref' => $base_metric_doc_ref );
Expand All @@ -724,7 +788,7 @@ foreach my $job_args (@jobs) {
push(@pids, $pid);
$num_jobs++;
} else {
my $num_metric_docs_submitted = index_metrics('index', $$job_args{'tool-file'}, $$job_args{'collector'}, $$job_args{'num'}, $$job_args{'doc-ref'});
my $num_metric_docs_submitted = index_metrics('index', $$job_args{'tool-dir'}, $$job_args{'tool-file'}, $$job_args{'collector'}, $$job_args{'num'}, $$job_args{'doc-ref'});
#$tool_dir . "/" . $tool_file, $collector, $num, $base_metric_doc_ref);
exit 0;
}
Expand Down Expand Up @@ -880,11 +944,12 @@ if (exists $result{'iterations'}) {
for (my $j = 0; $j < scalar(@{ $data{'periods'}[$k]{'metric-files'} }); $j++) {
# Metric data is still in other file(s). For each member in 'metric-files' array,
# there should be a 2 files with the same prefix
my $metric_file_prefix = $run_dir . "/" . $cs_id_dir . "/" . $data{'periods'}[$k]{'metric-files'}[$j];
my $metric_file_prefix = $data{'periods'}[$k]{'metric-files'}[$j];
my $metric_dir = $run_dir . "/" . $cs_id_dir;
my $this_begin;
my $this_end;
# index_metric() to return the easliest-begin and latest-end for metric types matching the primary-metric
(my $num_metric_docs_submitted, $this_begin, $this_end) = index_metrics('queue', $metric_file_prefix, $cs_name, $cs_id, $base_metric_doc_ref, $data{'benchmark'}, $data{'primary-metric'});
(my $num_metric_docs_submitted, $this_begin, $this_end) = index_metrics('queue', $metric_dir, $metric_file_prefix, $cs_name, $cs_id, $base_metric_doc_ref, $data{'benchmark'}, $data{'primary-metric'});
# From processing all metric files, get the very-earliest-begin and very-latest-end
if (defined $this_begin and defined $this_end) {
if (not defined $earliest_begin or $earliest_begin > $this_begin) {
Expand Down

0 comments on commit 3125528

Please sign in to comment.