Observability icons

The 68 observability icons in DevOps, described

Alertalert

An alert — something fired, a notification raised or a page sent out.

Alert rulealert-rule

An alert rule — the condition that has to be true before an alert fires.

Allocation rateallocation-rate

A machine frame with a rising trend — how fast memory is being allocated.

Auto instrumentauto-instrument

A bracket pair with two span lines already laid in — instrumentation that arrives without an edit.

Baggagebaggage

A span bar carrying a small case — key-values that ride along with every hop of a trace.

Brownoutbrownout

A brownout — still on, just less of it, a service degraded to stay up.

Cold start tracecold-start-trace

A span bar with a tick and a gap before the work — the time a cold start ate.

Context propagationcontext-propagation

A span bar with an arrow running through it — trace context handed on to the next service.

Continuous profilecontinuous-profile

A machine frame with a pulse running through it — profiling that never stops.

Correlation IDcorrelation-id

A span bar with an equals sign — the one id that says two records belong to the same request.

CPU profilecpu-profile

A machine frame with a stack of narrowing bars — where the processor's time went.

Debug symboldebug-symbol

A page with a tag on it — the names that turn addresses back into code.

Dependency mapdependency-map

Two nodes leading down into one — what a service needs before it can answer.

Error budgeterror-budget

An error budget — how much failure an SLO allows you before you have to stop shipping.

Exemplarexemplar

A span bar with a pin in it — one real trace picked to stand for a metric's bucket.

Flame graphflamegraph

A flame graph — where the time actually went, stack by stack in a profile.

Goroutine dumpgoroutine-dump

A scatter of small points landing in a tray — every goroutine and what it was doing.

Head samplinghead-sampling

A span bar with a funnel at its head — the keep-or-drop decision made before the trace starts.

Health checkhealth-check

A health check — the probe that asks whether a service is still alive and answering.

Heap dumpheap-dump

A pile of objects landing in a tray — the whole heap written out to look at.

Instrument codeinstrument-code

A bracket pair with a span bar laid inside — a span put into the code by hand.

Latency trendlatency-trend

A timer dial with a rising line on its face — how response time is trending.

Live taillive-tail

A receipt whose last line is still being written — logs followed as they arrive.

Load sheddingload-shed

Load shedding — better to drop some requests than sink the whole service.

Lock contentionlock-contention

A machine frame with a lock — threads waiting on the same mutex.

Loglog

A log — what happened, written down as it happened, entry by entry.

Log fieldlog-field

A receipt with one field marked on the middle line — a single key pulled out of a log line.

Log indexlog-index-obs

A receipt with a short list — logs indexed so a query can find them.

Log levellog-level

A receipt with three lines that step up in length — debug, info, warn, error.

Log parselog-parse

A receipt with a line and an arrow breaking down from it — a raw line split into its parts.

Log querylog-query

A receipt with a lens — a search run across logs.

Log retentionlog-retention

A receipt with a clock — how long logs are kept before they are dropped.

Log rotationlog-rotation

A receipt with an up and a down chevron — the current file swapped out for a fresh one.

Log samplinglog-sampling

A receipt with a funnel — only some log lines kept.

Log shipperlog-shipper

A receipt with an arrow across it — logs forwarded off the host to wherever they are kept.

Log streamlog-stream

A log stream — lines arriving live as they happen, tailed and followed.

Log volumelog-volume

A receipt with three rising bars — how much log a service writes.

Memory profilememory-profile

A machine frame with three bars standing in it — where the memory went.

Observabilityobservability

Observability — being able to see what a system is doing through logs, metrics and traces.

OTel collectorotel-collector

Three streams meeting in one and landing in a tray — telemetry gathered before it goes anywhere.

OTel exporterotel-exporter

An arrow rising out of a tray — telemetry sent on to a backend.

Packet losspacket-loss

Packet loss — one of the packets never arrived, a gap in the network stream.

Parent spanparent-span

A span bar with a wide bar over an indented one — the span this span was started from.

Profile diffprofile-diff

A machine frame with a plus over a minus — two profiles compared.

Queue depthqueue-depth

Queue depth — how much work is waiting right now, the backlog height.

Redact logredact-log

A receipt with a line blacked out — a value masked before the log leaves.

Request timelinerequest-timeline

A span bar with three ticks — the moments of one request in order.

Root spanroot-span

A span bar with a wide bar and a bracket of children under it — the span with no parent.

Sample logsample-log

A receipt with a point between two lines — one log picked out of many.

Service mapservice-map

Four nodes joined corner to corner — which services talk to which.

SLAsla

An SLA — the service level agreement, a promise with a signature on it.

SLIsli

An SLI — the service level indicator you actually measure, such as latency or error rate.

SLOslo

An SLO — the service level objective you hold yourself to, the target for an indicator.

Span attributespan-attribute

A span bar with a short list — the key-values attached to a span.

Span durationspan-duration

A span bar with a clock — how long the span took.

Span eventspan-event

A span bar with one point marked on it — something that happened at an instant inside the span.

Span IDspan-id

A span bar with a hash in it — the id of this one span.

Span linkspan-link

A span bar with two points joined — a span that refers to another trace's span.

Span statusspan-status

A span bar with a check — whether the span ended ok or in error.

Structured logstructured-log

A receipt whose lines are split into key and value — a log written as fields, not prose.

Tail samplingtail-sampling

A span bar with a funnel at its tail — the keep-or-drop decision made once the trace is complete.

Thread dumpthread-dump

Three threads landing in a tray — every thread's stack at one instant.

Tracetrace

A trace — one request followed through every service it touched, span by span.

Trace contexttrace-context

A span bar holding angle brackets — the headers that carry a trace across a request.

Trace IDtrace-id

A span bar with a tag — the id that names the whole trace.

Trace samplingtrace-sampling

A span bar with a funnel — only some traces kept.

Trace waterfalltrace-waterfall

A span bar with three staggered bars inside — the spans of one trace laid out in time.

Uptimeuptime

Uptime — how long a service has stayed up and available without an outage.

In code, each is one import — import { Alert } from "@iconmind/react/icons/alert" — and the same name in Vue, Svelte, Solid, Preact, React Native, Astro, Blade and Flutter.

Narrow it by tag