Skip to content

TRITON-2516 optimize cpu_detailed cmon gz plugin #3

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 2 commits into from
Oct 29, 2025
Merged

TRITON-2516 optimize cpu_detailed cmon gz plugin #3

Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
909dbd4
increase timeout for cpu_detailed plugin
jsadowyj Oct 28, 2025
a50b4ea
optimize cpu_detailed plugin with awk
jsadowyj Oct 28, 2025
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
170 changes: 106 additions & 64 deletions gz-plugins/cpu_detailed.prom
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -30,67 +30,109 @@ if [ -z "$cpu_stats" ]; then
exit 0
fi

# Get list of CPU IDs
cpu_ids=$(psrinfo | cut -f1)

# Parse cpu_stats once into associative arrays
declare -A user_sec
declare -A kernel_sec
declare -A idle_sec
declare -A intr_sec

while IFS=$'\t' read -r key value; do
if [[ $key =~ ^cpu:([0-9]+):sys:cpu_nsec_user$ ]]; then
value_sec=$(awk "BEGIN {printf \"%.9f\", $value / 1000000000}")
user_sec[${BASH_REMATCH[1]}]=$value_sec
elif [[ $key =~ ^cpu:([0-9]+):sys:cpu_nsec_kernel$ ]]; then
value_sec=$(awk "BEGIN {printf \"%.9f\", $value / 1000000000}")
kernel_sec[${BASH_REMATCH[1]}]=$value_sec
elif [[ $key =~ ^cpu:([0-9]+):sys:cpu_nsec_idle$ ]]; then
value_sec=$(awk "BEGIN {printf \"%.9f\", $value / 1000000000}")
idle_sec[${BASH_REMATCH[1]}]=$value_sec
elif [[ $key =~ ^cpu:([0-9]+):sys:cpu_nsec_intr$ ]]; then
value_sec=$(awk "BEGIN {printf \"%.9f\", $value / 1000000000}")
intr_sec[${BASH_REMATCH[1]}]=$value_sec
fi
done <<< "$cpu_stats"

# Set TTL to 0 seconds to ensure accurate rate calculations
printf '# OPTION ttl 0\n'

# Output metric headers
printf '# HELP cn_cpu_core_user_seconds_total CPU time spent in user mode per core\n'
printf '# TYPE cn_cpu_core_user_seconds_total counter\n'
printf '# HELP cn_cpu_core_system_seconds_total CPU time spent in system mode per core\n'
printf '# TYPE cn_cpu_core_system_seconds_total counter\n'
printf '# HELP cn_cpu_core_idle_seconds_total CPU time spent idle per core\n'
printf '# TYPE cn_cpu_core_idle_seconds_total counter\n'
printf '# HELP cn_cpu_core_interrupt_seconds_total CPU time spent servicing interrupts per core\n'
printf '# TYPE cn_cpu_core_interrupt_seconds_total counter\n'

# Loop through CPUs and output all metrics using array lookups
for cpu_id in $cpu_ids; do
if [ -n "${user_sec[$cpu_id]}" ]; then
printf 'cn_cpu_core_user_seconds_total{core="%s"}\t%s\n' "$cpu_id" "${user_sec[$cpu_id]}"
fi
if [ -n "${kernel_sec[$cpu_id]}" ]; then
printf 'cn_cpu_core_system_seconds_total{core="%s"}\t%s\n' "$cpu_id" "${kernel_sec[$cpu_id]}"
fi
if [ -n "${idle_sec[$cpu_id]}" ]; then
printf 'cn_cpu_core_idle_seconds_total{core="%s"}\t%s\n' "$cpu_id" "${idle_sec[$cpu_id]}"
fi
if [ -n "${intr_sec[$cpu_id]}" ]; then
printf 'cn_cpu_core_interrupt_seconds_total{core="%s"}\t%s\n' "$cpu_id" "${intr_sec[$cpu_id]}"
fi
done

# Output CPU info metrics (only need to output once, use cpu 0)
brand=$(echo "$cpu_stats" | grep "^cpu_info:0:" | grep ':brand' | cut -f2-)
chip_id=$(echo "$cpu_stats" | grep "^cpu_info:0:" | grep ':chip_id' | awk '{print $2}')
clock_mhz=$(echo "$cpu_stats" | grep "^cpu_info:0:" | grep ':clock_MHz' | awk '{print $2}')

if [ -n "$brand" ] && [ -n "$chip_id" ] && [ -n "$clock_mhz" ]; then
printf '# HELP cn_cpu_info CPU information\n'
printf '# TYPE cn_cpu_info gauge\n'
printf 'cn_cpu_info{brand="%s",chip_id="%s",clock_mhz="%s"}\t1\n' "$brand" "$chip_id" "$clock_mhz"
fi
# Parse and output all metrics using a single awk script
echo "$cpu_stats" | awk '
BEGIN {
FS = "\t"
cpu_order_count = 0
}

function extract_cpu_id(key, prefix) {
sub(prefix, "", key)
return key
}

function shell_escape(s) {
gsub(/\\/, "\\\\", s)
gsub(/"/, "\\\"", s)
return s
}

function record_cpu_order(cpu_id) {
# Track CPU order - only add if we havent seen this CPU before
if (!(cpu_id in cpu_seen)) {
cpu_seen[cpu_id] = 1
cpu_order[++cpu_order_count] = cpu_id
}
}

/^cpu:[0-9]+:sys:cpu_nsec_user\t/ {
cpu_id = extract_cpu_id($1, "cpu:")
sub(/:sys:cpu_nsec_user$/, "", cpu_id)
record_cpu_order(cpu_id)
user_sec[cpu_id] = sprintf("%.9f", $2 / 1000000000)
}

/^cpu:[0-9]+:sys:cpu_nsec_kernel\t/ {
cpu_id = extract_cpu_id($1, "cpu:")
sub(/:sys:cpu_nsec_kernel$/, "", cpu_id)
record_cpu_order(cpu_id)
kernel_sec[cpu_id] = sprintf("%.9f", $2 / 1000000000)
}

/^cpu:[0-9]+:sys:cpu_nsec_idle\t/ {
cpu_id = extract_cpu_id($1, "cpu:")
sub(/:sys:cpu_nsec_idle$/, "", cpu_id)
record_cpu_order(cpu_id)
idle_sec[cpu_id] = sprintf("%.9f", $2 / 1000000000)
}

/^cpu:[0-9]+:sys:cpu_nsec_intr\t/ {
cpu_id = extract_cpu_id($1, "cpu:")
sub(/:sys:cpu_nsec_intr$/, "", cpu_id)
record_cpu_order(cpu_id)
intr_sec[cpu_id] = sprintf("%.9f", $2 / 1000000000)
}

# CPU info from cpu 0
/^cpu_info:0:/ && /:brand\t/ {
brand = shell_escape($2)
}

/^cpu_info:0:/ && /:chip_id\t/ {
chip_id = $2
}

/^cpu_info:0:/ && /:clock_MHz\t/ {
clock_mhz = $2
}

END {
# Print TTL option
printf "# OPTION ttl 0\n"

# Print metric headers
printf "# HELP cn_cpu_core_user_seconds_total CPU time spent in user mode per core\n"
printf "# TYPE cn_cpu_core_user_seconds_total counter\n"
printf "# HELP cn_cpu_core_system_seconds_total CPU time spent in system mode per core\n"
printf "# TYPE cn_cpu_core_system_seconds_total counter\n"
printf "# HELP cn_cpu_core_idle_seconds_total CPU time spent idle per core\n"
printf "# TYPE cn_cpu_core_idle_seconds_total counter\n"
printf "# HELP cn_cpu_core_interrupt_seconds_total CPU time spent servicing interrupts per core\n"
printf "# TYPE cn_cpu_core_interrupt_seconds_total counter\n"

# Output per-CPU metrics in the order from kstat
for (i = 1; i <= cpu_order_count; i++) {
cpu_id = cpu_order[i]

if (cpu_id in user_sec) {
printf "cn_cpu_core_user_seconds_total{core=\"%s\"}\t%s\n", cpu_id, user_sec[cpu_id]
}
if (cpu_id in kernel_sec) {
printf "cn_cpu_core_system_seconds_total{core=\"%s\"}\t%s\n", cpu_id, kernel_sec[cpu_id]
}
if (cpu_id in idle_sec) {
printf "cn_cpu_core_idle_seconds_total{core=\"%s\"}\t%s\n", cpu_id, idle_sec[cpu_id]
}
if (cpu_id in intr_sec) {
printf "cn_cpu_core_interrupt_seconds_total{core=\"%s\"}\t%s\n", cpu_id, intr_sec[cpu_id]
}
}

# Output CPU info metric
if (brand != "" && chip_id != "" && clock_mhz != "") {
printf "# HELP cn_cpu_info CPU information\n"
printf "# TYPE cn_cpu_info gauge\n"
printf "cn_cpu_info{brand=\"%s\",chip_id=\"%s\",clock_mhz=\"%s\"}\t1\n", brand, chip_id, clock_mhz
}
}'