Monitoring & Observability glossary

Every Monitoring & Observability word in this course, in plain words.

aggregation

Adding many series into fewer: sum, avg, max, count, with by (…) to keep some labels.

In BM lessons: aggregation (pengagregatan): Gabungkan banyak siri jadi sedikit: sum, avg, max, count, dengan by (…) untuk kekalkan label tertentu.

alert

A message that calls a person because something needs action now.

In BM lessons: amaran (alert): Mesej yang panggil seseorang kerana sesuatu perlu tindakan sekarang.

alert rule

A PromQL question Prometheus asks again and again. When the answer has rows, the control room's bell gets ready.

In BM lessons: alert rule: Soalan PromQL yang Prometheus tanya berulang kali. Bila jawapannya ada baris, loceng bilik kawalan bersedia.

Alertmanager

The control room's dispatcher. It takes bells from Prometheus, groups them, picks who to call and stays quiet when told.

In BM lessons: Alertmanager: Pengagih di bilik kawalan. Ia terima loceng dari Prometheus, kumpulkannya, pilih siapa dipanggil dan senyap bila disuruh.

blameless postmortem

A written look back after an incident. It asks what made the mistake easy, never who to blame.

In BM lessons: postmortem tanpa menuding jari: Catatan selepas insiden. Ia tanya apa yang buat kesilapan itu mudah berlaku, bukan siapa nak dipersalahkan.

burn rate

How fast you spend the error budget. 1 means it lasts exactly the SLO window; 14.4 means a month's budget is gone in about two days.

In BM lessons: burn rate: Berapa laju bajet ralat dibelanjakan. 1 bermaksud ia cukup tepat sepanjang tempoh SLO; 14.4 bermaksud bajet sebulan habis dalam lebih kurang dua hari.

cardinality

How many series a metric makes. Labels like user id or order id make far too many.

In BM lessons: kardinaliti: Berapa banyak siri yang satu metrik hasilkan. Label seperti id pengguna atau id pesanan buat terlalu banyak.

Collector

OpenTelemetry's sorting office. Receivers take data in, processors tidy it, exporters send it on, joined by pipelines.

In BM lessons: Collector: Pejabat sortir OpenTelemetry. Receiver terima data, processor kemaskan, exporter hantar keluar, disambung oleh pipeline.

context propagation

Passing the trace ID to the next service, in a traceparent header, so the order keeps its number.

In BM lessons: context propagation: Hantar trace ID ke servis seterusnya, dalam header traceparent, supaya pesanan kekal dengan nombornya.

counter

A number that only goes up, like orders served. It drops to 0 only when the app restarts.

In BM lessons: counter (pembilang): Nombor yang hanya naik, seperti pesanan siap. Ia jatuh ke 0 hanya bila aplikasi mula semula.

dashboard

A wall of graphs you look at when you choose. It never wakes anyone up.

In BM lessons: dashboard (papan pemuka): Dinding graf yang anda tengok bila anda mahu. Ia tak pernah kejutkan sesiapa.

dashboard variable

A drop-down at the top of a Grafana dashboard. Pick shop-2 and every panel's query swaps $instance for shop-2.

In BM lessons: dashboard variable: Senarai pilihan di atas dashboard Grafana. Pilih shop-2 dan query setiap panel tukar $instance kepada shop-2.

error budget

The bad you may spend: 100% minus the SLO. Spend it on change, not on surprises.

In BM lessons: bajet ralat (error budget): Ruang silap yang boleh anda guna: 100% tolak SLO. Guna untuk perubahan, bukan kejutan.

exemplar

A sample trace ID pinned to a metric point. Click the dot on the graph, open that very order's trace.

In BM lessons: exemplar: Contoh trace ID yang dilekat pada titik metrik. Klik titik pada graf, buka trace pesanan itu sendiri.

exporter

A translator: it reads numbers from something (a machine, a database) and shows them at /metrics.

In BM lessons: exporter: Penterjemah: ia baca nombor daripada sesuatu (mesin, pangkalan data) dan tunjuk di /metrics.

four golden signals

Latency, traffic, errors and saturation: the four screens to hang first.

In BM lessons: empat isyarat emas: Latency, traffic, errors dan saturation: empat skrin yang patut dipasang dulu.

gauge

A number that goes up and down, like customers waiting or free memory.

In BM lessons: gauge (meter): Nombor yang naik dan turun, seperti pelanggan menunggu atau memori kosong.

Grafana

The big screens: it draws dashboards from data kept elsewhere, like Prometheus.

In BM lessons: Grafana: Skrin besar: ia lukis dashboard daripada data yang disimpan di tempat lain, seperti Prometheus.

grouping

Alertmanager puts alerts with the same labels (like alertname and job) into one message, so ten bells make one call.

In BM lessons: grouping (kumpulan): Alertmanager satukan amaran yang sama label (contohnya alertname dan job) dalam satu mesej, jadi sepuluh loceng jadi satu panggilan.

histogram

Counts requests into time buckets (le = less or equal), plus a _sum and a _count.

In BM lessons: histogram: Kira permintaan ke dalam baldi masa (le = kurang atau sama), serta _sum dan _count.

incident commander

The one person who runs an incident: sets the plan, gives jobs and decides. They don't fix things themselves.

In BM lessons: incident commander: Seorang yang mengetuai insiden: tetapkan pelan, beri tugas dan buat keputusan. Dia tidak baiki sendiri.

inhibition

While a big alert fires (the whole site is down), Alertmanager mutes the smaller alerts it causes.

In BM lessons: inhibition: Semasa amaran besar berbunyi (seluruh laman mati), Alertmanager senyapkan amaran kecil yang berpunca daripadanya.

instant vector

One value per series, at one moment. Graphs draw these.

In BM lessons: instant vector (vektor instant): Satu nilai untuk setiap siri, pada satu saat. Graf lukis jenis ini.

label

A sticky note on a series, like stall="cendol". Every new combination makes a new series.

In BM lessons: label: Nota lekat pada siri, seperti stall="cendol". Setiap gabungan baru jadi siri baru.

latency

How long one request takes, from asking to answer.

In BM lessons: latency (kependaman): Berapa lama satu permintaan ambil, dari tanya hingga dapat jawapan.

log

A diary line from the app: what happened, when, with details.

In BM lessons: log: Satu baris diari daripada aplikasi: apa yang berlaku, bila, dengan butiran.

log stream

One diary in Loki: all lines with the same set of labels. Too many label values make too many diaries.

In BM lessons: log stream: Satu diari dalam Loki: semua baris dengan set label yang sama. Terlalu banyak nilai label buat terlalu banyak diari.

LogQL

Loki's query language. Pick diaries by label in braces, then filter lines: {app="kedai-api"} |= "timeout".

In BM lessons: LogQL: Bahasa query Loki. Pilih diari ikut label dalam kurungan, kemudian tapis baris: {app="kedai-api"} |= "timeout".

Loki

Grafana's log store. It indexes only the labels of each diary, not every word, so it stays small and cheap.

In BM lessons: Loki: Stor log Grafana. Ia indeks label setiap diari sahaja, bukan setiap perkataan, jadi ia kekal kecil dan murah.

metric

A number measured again and again over time, like orders per minute.

In BM lessons: metrik: Nombor yang diukur berulang kali dari masa ke masa, contohnya pesanan seminit.

monitoring

Watching known numbers on the control room screens, with an alarm when one goes bad.

In BM lessons: pemantauan (monitoring): Pantau nombor yang dah dikenal pada skrin bilik kawalan, dengan penggera bila satu jadi teruk.

observability

How well you can ask new questions about a system from the outside, using what it tells you.

In BM lessons: observability (kebolehcerapan): Sejauh mana anda boleh tanya soalan baru tentang sistem dari luar, guna apa yang ia beritahu.

on call

Your turn to answer alerts, for a set time, with a plan and a backup person.

In BM lessons: on-call: Giliran anda jawab amaran, untuk satu tempoh, dengan pelan dan orang sandaran.

OpenTelemetry (OTel)

One shared way to make and send metrics, logs and traces. Not a database.

In BM lessons: OpenTelemetry (OTel): Satu cara bersama untuk hasilkan dan hantar metrik, log dan trace. Bukan pangkalan data.

OpenTelemetry SDK

The library inside your app that writes down metrics, logs and spans, then sends them out with OTLP.

In BM lessons: OpenTelemetry SDK: Pustaka dalam aplikasi anda yang catat metrik, log dan span, kemudian hantarnya keluar dengan OTLP.

OTLP

The OpenTelemetry Protocol: one shared way to send metrics, logs and traces. Usually port 4317 (gRPC) or 4318 (HTTP).

In BM lessons: OTLP: OpenTelemetry Protocol: satu cara sama untuk hantar metrik, log dan trace. Selalunya port 4317 (gRPC) atau 4318 (HTTP).

panel

One screen on the control room wall. Each panel runs one query and draws it: a line graph, a big number, a gauge or a table.

In BM lessons: panel: Satu skrin di dinding bilik kawalan. Setiap panel jalankan satu query dan lukisnya: graf garis, nombor besar, gauge atau jadual.

pending

The bell is ready but still waiting. The problem must last the whole `for` time before it rings (firing).

In BM lessons: pending: Loceng dah sedia tapi masih menunggu. Masalah mesti berterusan sepanjang masa `for` sebelum ia berbunyi (firing).

Prometheus

The control room's clerk: it collects numbers on a timer, keeps them, and answers PromQL questions.

In BM lessons: Prometheus: Kerani bilik kawalan: ia kutip nombor ikut jadual, simpan, dan jawab soalan PromQL.

PromQL

The language you use to ask Prometheus questions.

In BM lessons: PromQL: Bahasa untuk tanya soalan kepada Prometheus.

quantile (p95)

p95 = 95% of requests were this fast or faster. The slowest 5% took longer.

In BM lessons: kuantil (p95): p95 = 95% permintaan selaju ini atau lebih laju. 5% yang paling lambat ambil masa lebih.

range vector

Every sample per series in a window, like [5m]. Functions like rate() need one.

In BM lessons: range vector (vektor julat): Semua sampel setiap siri dalam satu tetingkap, seperti [5m]. Fungsi seperti rate() perlukannya.

rate()

How fast a counter grows, per second, averaged over a window. It handles restarts.

In BM lessons: rate(): Berapa laju counter naik, sesaat, dipuratakan dalam satu tetingkap. Ia faham bila aplikasi mula semula.

recording rule

A saved question: Prometheus runs it on a timer and keeps the answer as a new series.

In BM lessons: recording rule: Soalan yang disimpan: Prometheus jalankannya ikut jadual dan simpan jawapan sebagai siri baru.

RED

Rate, Errors, Duration: three questions for every service.

In BM lessons: RED: Rate, Errors, Duration: tiga soalan untuk setiap servis.

route

A rule in Alertmanager that sends an alert to a receiver, by its labels. The first match wins.

In BM lessons: route (hala): Peraturan dalam Alertmanager yang hantar amaran ke penerima, ikut labelnya. Padanan pertama menang.

runbook

The fix-it card next to the bell: what this alert means, what to check first and how to make it safe.

In BM lessons: runbook: Kad baiki di sebelah loceng: maksud amaran ini, apa nak semak dulu dan cara jadikannya selamat.

saturation

How full the system is: a growing queue, a busy stove, a nearly full disk.

In BM lessons: saturation (ketepuan): Sejauh mana sistem dah penuh: barisan makin panjang, dapur sibuk, cakera hampir penuh.

scrape

Prometheus visits a target on a timer and copies its numbers from /metrics.

In BM lessons: scrape (kutip): Prometheus lawat target ikut jadual dan salin nombornya dari /metrics.

silence

A note on the dispatcher's desk: don't call anyone for these alerts until a set time. It has an author and a reason.

In BM lessons: silence (senyap): Nota atas meja pengagih: jangan panggil sesiapa untuk amaran ini hingga masa tertentu. Ia ada penulis dan sebab.

SLI (service level indicator)

A number that shows how users feel the service, like the share of good checkouts.

In BM lessons: SLI (penunjuk tahap servis): Nombor yang tunjuk apa pengguna rasa tentang servis, contohnya peratus bayaran yang berjaya.

SLO (service level objective)

The target for an SLI over a time window, like 99.9% good checkouts in 30 days.

In BM lessons: SLO (objektif tahap servis): Sasaran untuk SLI dalam satu tempoh, contohnya 99.9% bayaran berjaya dalam 30 hari.

span

One step of an order, with a start and an end time. Spans nest: a parent span holds child spans.

In BM lessons: span: Satu langkah pesanan, dengan masa mula dan tamat. Span bersarang: span induk ada span anak.

structured log

A diary line with named fields, like JSON: level, msg, order_id. Computers can filter it; people can still read it.

In BM lessons: log berstruktur: Baris diari dengan medan bernama, macam JSON: level, msg, order_id. Komputer boleh tapis; manusia masih boleh baca.

summary

Quantiles worked out inside the app. You cannot add them up across servers.

In BM lessons: summary: Kuantil yang dikira dalam aplikasi. Anda tak boleh jumlahkan merentas server.

symptom

What the customer feels: failed orders, slow pages. Alert on symptoms; look at causes (CPU, disk) on the screens.

In BM lessons: simptom: Apa yang pelanggan rasa: pesanan gagal, laman lambat. Beri amaran pada simptom; tengok punca (CPU, cakera) di skrin.

target

One address Prometheus scrapes, like node-exporter:9100. Its job and instance become labels.

In BM lessons: target: Satu alamat yang Prometheus scrape, seperti node-exporter:9100. Job dan instance-nya jadi label.

time series

One metric name plus one set of labels: a line of values over time.

In BM lessons: time series (siri masa): Satu nama metrik dengan satu set label: satu garisan nilai dari masa ke masa.

trace

One order followed through every service, with the time each step took.

In BM lessons: trace (jejak): Satu pesanan dijejak melalui setiap servis, dengan masa setiap langkah.

trace ID

The order number every span and log line of one request shares: 32 hex characters.

In BM lessons: trace ID: Nombor pesanan yang dikongsi setiap span dan baris log satu permintaan: 32 aksara hex.

USE

Utilization, Saturation, Errors: three questions for every machine part, like CPU or disk.

In BM lessons: USE: Utilization, Saturation, Errors: tiga soalan untuk setiap bahagian mesin, seperti CPU atau cakera.