Glossary
One-line definitions of the terms used in FuseScope and in these pages, in alphabetical order. For longer explanations with examples, see Core concepts.
Aggregation
Section titled “Aggregation”A calculation over many spans, logs or data points that returns a number, such as
count(), avg(duration_nano) or p99(duration_nano). See
Query builder.
Alert rule
Section titled “Alert rule”A query FuseScope runs on a schedule, with a threshold. When the result crosses the threshold, an alert fires and a notification is sent. See Alert rules.
A score from 0 to 1 for how satisfied users are with a service’s response times, based on a target time T: requests within T count fully, within 4 × T half, and slower or failed ones not at all. See Core concepts and Apdex threshold.
Attribute
Section titled “Attribute”A key and value attached to a span, log or data point, such as http.method = GET.
Used to filter and group data.
Auth domain
Section titled “Auth domain”An email domain, such as example.com, tied to a single sign-on provider. Users with an
email in that domain sign in through the provider. See
Single sign-on.
Breadcrumbs
Section titled “Breadcrumbs”In Real User Monitoring, the views, clicks and requests of a session in the 2 minutes before a JavaScript error, so you can see what the user did just before it. See Error detail.
Cardinality
Section titled “Cardinality”The number of different values a key has. http.method has low cardinality (a
handful of values); user.id has high cardinality. Grouping by a high-cardinality key
produces very many series.
Celery worker
Section titled “Celery worker”A process that runs Celery tasks, Python jobs queued by your code to run in the background. See Celery.
ClickHouse
Section titled “ClickHouse”The database FuseScope stores traces, logs and metrics in.
Collector
Section titled “Collector”The OpenTelemetry Collector: a program that receives telemetry from your applications,
processes it and writes it to ClickHouse. FuseScope ships its own build. Next to your
applications you can run the upstream otelcol-contrib build as an agent, with
receivers and exporters of its own. See
Architecture and
Collector on Docker.
Consumer lag
Section titled “Consumer lag”How many messages a Kafka consumer group still has to read in a partition. Lag that keeps growing means the consumers can’t keep up. See Messaging Queues.
Counter
Section titled “Counter”A Sum metric that only goes up, such as requests served. Queried with a rate. See Metric basics.
Dashboard
Section titled “Dashboard”A page of panels (charts, tables, numbers) you arrange yourself. See Dashboards.
Dashboard variable
Section titled “Dashboard variable”A control at the top of a dashboard, such as a list of services, whose value the
panel queries use as $name. Changing it changes what every panel that uses it shows.
See Dashboard variables.
Data source catalog
Section titled “Data source catalog”The Get Started page, organised by goal, that lists the languages, log sources, infrastructure and integrations you can send data from, each with its setup page. See Data source catalog.
Downstream
Section titled “Downstream”The services, databases and external addresses a service calls. See Upstream and downstream.
Duration
Section titled “Duration”How long a span took, from start to end. Stored in nanoseconds as duration_nano.
Endpoint
Section titled “Endpoint”One kind of request or call a service handles, shown as a row of the
Endpoints tab: an
operation, or an http.route or rpc.method value.
Entrypoint span
Section titled “Entrypoint span”The first span of a request inside one service, where the request entered it. Service latency and error rate are computed from these spans.
Error group
Section titled “Error group”In Real User Monitoring, all the JavaScript errors with the same fingerprint: one bug, however many times and in however many browsers it happened. See Errors.
Error rate
Section titled “Error rate”The share of requests that failed, as a percentage.
Evaluation window
Section titled “Evaluation window”How much recent data an alert rule looks at each time it runs, such as the last 5 minutes. See Evaluation window and frequency.
Exception
Section titled “Exception”An error your code raised, recorded on a span as an exception event with its type,
message and stack trace. See Exceptions.
Exponential histogram
Section titled “Exponential histogram”A histogram whose buckets grow exponentially in size, so it covers a wide range of values without bucket settings.
Exporter
Section titled “Exporter”A collector component that sends data on, such as otlp (gRPC) or otlphttp to
FuseScope, or debug to the collector’s own output. See
OTLP.
External domain
Section titled “External domain”A host your services call over HTTP, such as api.stripe.com or 10.0.4.11, taken
from the URL of their client spans. It doesn’t need to send telemetry itself. See
External APIs.
Filter
Section titled “Filter”A condition that selects which data a query uses, such as
service.name = 'driver'. See Search syntax.
Fingerprint
Section titled “Fingerprint”A short id that groups JavaScript errors, computed from the exception type, the message with its numbers, ids and URLs removed, and the file and function of the top stack frame. See How errors are grouped.
Flame graph
Section titled “Flame graph”A picture of a trace where time runs left to right and each span sits under the span that called it. See Trace details.
A metric type for a current value that goes up and down, such as memory in use. See Metric types.
GenAI semantic conventions
Section titled “GenAI semantic conventions”The OpenTelemetry attribute names for LLM, tool and agent calls, such as
gen_ai.request.model and gen_ai.usage.input_tokens. AI Observability reads them. See
AI Observability for Python.
Group by
Section titled “Group by”Splitting a query’s result into one series or row per value of a key, such as one
line per service.name.
Highlight
Section titled “Highlight”A change FuseScope flags on a service because it crossed a rule: an error rate of 50% or more, or the error rate, p99 latency or request rate at least doubling against the previous period. On the service page, also a downstream dependency that fails half its calls. See Highlight badges.
Histogram
Section titled “Histogram”A metric type that counts how many values fall into each range (bucket), used to compute percentiles.
Impersonation mode
Section titled “Impersonation mode”A setting that turns sign-in off: every request acts as the root user, an admin. For trusted local or demo instances only. See Impersonation mode.
Inferred dependency
Section titled “Inferred dependency”Something your services call that sends no traces itself, such as a managed database,
a queue or a third-party API. FuseScope draws it on the
Service Map from the callers’
client spans (db.system, messaging.system, peer.service or server.address),
with a dashed outline.
Instrumentation
Section titled “Instrumentation”The code or agent that makes an application produce traces, logs and metrics. See Instrument applications.
Integration
Section titled “Integration”A ready-made setup for one product, such as PostgreSQL or Redis: a collector configuration, dashboards and a connection check. See Integrations.
Latency
Section titled “Latency”How long a request takes to get a response.
LLM tokens
Section titled “LLM tokens”The pieces of text a language model reads (input tokens) and writes (output tokens), roughly three quarters of a word each. Providers charge per million tokens. See AI Observability.
A timestamped record a program writes, with a body (the message) and a severity.
Log pipeline
Section titled “Log pipeline”A filter and a list of processors that change logs before they are stored, such as parsing JSON or setting the severity. See Log pipelines.
Meter signal
Section titled “Meter signal”A count of what the workspace ingested: log records, spans and metric data points,
and the size of logs and spans, written once per hour as signoz.meter.* metrics.
See Cost Meter.
Metric
Section titled “Metric”A named number measured over time, such as requests per second.
Monotonic
Section titled “Monotonic”Said of a counter that only goes up, such as total requests served.
Multiline log
Section titled “Multiline log”A log entry that spans several lines, such as a Java stack trace. A collector joins the lines into one record with a pattern that matches the first line. See Logs from files.
Node and pod
Section titled “Node and pod”In Kubernetes, a node is a machine of the cluster and a pod is one or more containers scheduled together on a node. See Infrastructure.
Notification channel
Section titled “Notification channel”A destination for alert notifications: Slack, email, PagerDuty, a webhook and others. See Notification channels.
OpenTelemetry
Section titled “OpenTelemetry”An open standard and set of libraries for producing traces, logs and metrics. Often shortened to OTel. FuseScope accepts data from any OpenTelemetry SDK.
Operation
Section titled “Operation”The name of a span, such as /dispatch, HTTP GET or GetDriver. Spans with the
same name in one service are the same operation.
The OpenTelemetry protocol, used to send telemetry over gRPC (port 4317) or HTTP (port 4318). See OTLP.
Page load
Section titled “Page load”In Real User Monitoring, the view a browser loaded the page on: the first view of a visit, or a reload. Only page loads have Web Vitals. The other views of a single-page app are route changes. See Page loads and route changes.
One chart, table, number, list or note on a dashboard, with its own query and title. See Dashboards.
Panel editor
Section titled “Panel editor”The full-screen page where you build a panel: its queries on the left, a preview in the middle and its settings on the right. See Panel editor.
Panel type
Section titled “Panel type”How a panel draws its data, such as Time Series, Bar Chart, Table or Text. Each type accepts some kinds of data and queries. See Panel types.
Partition
Section titled “Partition”One of the ordered parts a Kafka topic is split into, numbered from 0, so several
consumers can read the topic in parallel. See
Messaging Queues.
Percentile
Section titled “Percentile”The value below which a given share of measurements fall. P50 is the median; P95 and P99 show the slowest 5% and 1% of requests. See Core concepts.
Planned downtime
Section titled “Planned downtime”A maintenance window during which alert notifications are held back. Rules still run and record alerts. See Planned downtime.
Processor
Section titled “Processor”One step of a log pipeline, such as a JSON parser or a Severity Parser. A pipeline runs its processors in order. See Log pipelines.
Public key (RUM)
Section titled “Public key (RUM)”A write-only key, starting with rum_pub_, that a web app sends with its Real User
Monitoring data. It is safe to ship in a JavaScript bundle: it can only send data for its
application, from its allowed origins. See RUM keys.
A request for data: a source (traces, logs or metrics), a filter, and optionally an aggregation and group by.
How fast a counter grows, per second, such as 35 requests per second. See Why counters are queried with a rate.
Real User Monitoring (RUM)
Section titled “Real User Monitoring (RUM)”Measuring what real users experience in their browser: page loads, Web Vitals, JavaScript errors and the requests each page makes, linked to backend traces. See Real User Monitoring.
Receiver
Section titled “Receiver”A collector component that gets data in, by listening (otlp, syslog,
fluentforward) or by reading and scraping (filelog, hostmetrics, prometheus,
postgresql).
RED metrics
Section titled “RED metrics”Rate, errors and duration: how many requests a service handles, how many fail and how long they take. See Services.
Release
Section titled “Release”The version of your app that sent the data, the version passed to the RUM SDK’s
init() and stored as service.version. RUM screens filter by it, and source maps are
uploaded per release. See Source maps.
Requests and limits
Section titled “Requests and limits”The CPU and memory Kubernetes reserves for a container (request) and the most it may use (limit). Over its memory limit a container is killed; over its CPU limit it is slowed down. See Infrastructure.
Resource attribute
Section titled “Resource attribute”An attribute that describes what sent the data, such as service.name or
k8s.pod.name. It is the same on everything one program sends.
Root cause
Section titled “Root cause”The service where a failure starts, as opposed to the services that fail because they call it. The Service Map names a likely root cause: the deepest service with 5% errors or more whose callees are all healthy.
Root span
Section titled “Root span”The first span of a trace, the one with no parent. Its duration is the duration of the whole request.
Route change
Section titled “Route change”In Real User Monitoring, a view reached by navigating inside a single-page app, without reloading the page. Route changes have no Web Vitals but count time on view, errors and requests. See Views.
Route template
Section titled “Route template”A page path with its variable parts replaced by names, such as /product/:id for
/product/1234. RUM groups views and vitals by route template. See
Real User Monitoring.
Routing policy
Section titled “Routing policy”A rule that sends every alert whose labels match an expression, such as
service.name == "payment", to chosen notification channels. See
Routing policies.
Security Assertion Markup Language 2.0, an older single sign-on standard: the identity provider sends FuseScope a signed assertion with the user’s email. See SAML 2.0.
Sample
Section titled “Sample”One value of a metric at one moment, also called a data point. The Metrics explorer counts them per metric. See Samples and time series.
Scraper
Section titled “Scraper”A part of the hostmetrics receiver that reads one kind of host metric, such as
cpu, memory or filesystem. See Hosts.
Section
Section titled “Section”A titled group of panels on a dashboard. Sections can be folded, reordered and cloned. See Sections.
Service
Section titled “Service”One program that sends telemetry, named by its service.name resource attribute.
Session
Section titled “Session”In Real User Monitoring, one user’s visit to a web app, shared by all its tabs. It ends
after 15 minutes without activity or after 4 hours; the next session carries its id as
session.previous_id. Every RUM span and event carries its session.id. See
Sessions.
Severity
Section titled “Severity”How serious a log is, stored as severity_number (1 to 24) and severity_text.
FuseScope shows it as one of TRACE, DEBUG, INFO, WARN, ERROR, FATAL or UNKNOWN. See
Severity levels.
Single sign-on
Section titled “Single sign-on”Signing in to FuseScope with a company account from an identity provider, over OpenID Connect, Google Workspace or SAML, instead of a FuseScope password. See Single sign-on.
Source map
Section titled “Source map”A file your bundler writes next to each minified JavaScript bundle, mapping its positions back to your original files, functions and lines. Uploaded to FuseScope, it turns browser stack traces back into your code. See Source maps.
One piece of work inside a trace, such as an HTTP request or a database query, with a name, start time, duration and status.
Span event
Section titled “Span event”Something that happened at one moment during a span, such as an exception.
Stack trace
Section titled “Stack trace”The list of function calls that were running when an exception was raised, innermost first. It shows which line failed and how the program got there. See Read a stack trace.
A metric type for a total, such as requests served. Also called a counter when it is monotonic.
Summary (metric type)
Section titled “Summary (metric type)”A metric type that carries percentiles the application computed itself, sent by older Prometheus clients.
Symbolication
Section titled “Symbolication”Translating a minified stack trace back to the original source with a source map. FuseScope does it when you read an error, so maps uploaded later also fix past errors. See Source maps.
Temporality
Section titled “Temporality”How a counter reports its value. Cumulative: the total since the program started. Delta: the change since the last report. See Metric types and temporality.
Threshold
Section titled “Threshold”The value an alert rule compares its query with, such as 5 for a 5% error rate.
A rule can have several, named critical, warning and info. See
Thresholds and severity.
Throughput
Section titled “Throughput”How many requests a service handles per second, shown as the rate in requests per second (req/s).
Time on view
Section titled “Time on view”In Real User Monitoring, how long a user stayed on a view: from when it opened until the next navigation or until the tab was hidden. The Views tab shows the median. See Views.
Time range
Section titled “Time range”The period a page or query covers, such as the last 30 minutes.
Time series
Section titled “Time series”The values of one metric for one combination of label values, over time. See Samples and time series.
Every span of one request, linked into a tree by parent and child. All spans of a trace share a trace ID.
Trace ID
Section titled “Trace ID”The identifier shared by every span of a trace. Logs written during a request carry it too, which links them to the trace.
Time to first token: how long a streaming LLM call takes before the first words of the answer arrive. See AI Observability.
Upstream and downstream
Section titled “Upstream and downstream”Directions of calls between services. Upstream services call this one;
downstream ones are called by it. In HotROD, frontend is upstream of driver,
and redis-manual is downstream of it. See
Dependencies.
In Real User Monitoring, one page or route a user saw, either a page load
or a route change. Each view is a span named view, whose duration is
the time on view. See Views.
Waterfall
Section titled “Waterfall”A trace drawn as a tree of spans, one row each, with a bar for when each span ran. See Trace details.
Web Vitals
Section titled “Web Vitals”Google’s measures of page experience: LCP (loading), INP (responsiveness) and CLS (visual stability), each rated good, needs improvement or poor. See Web Vitals in plain words.
Zero-code instrumentation
Section titled “Zero-code instrumentation”An agent or launcher that instruments common frameworks and libraries without
changing your code, such as the OpenTelemetry Java agent or opentelemetry-instrument
for Python. See Instrument applications.