counterweight
f6656b0ff7
Asked to un-hide endpoint properties; the answer is that nothing was hidden.
All six hide-* options (hide-hostname, hide-url, hide-port, hide-conditions,
hide-errors, and dont-resolve-failed-conditions) already default to false
upstream, so hostname, URL, port, conditions and errors were all being shown.
The one setting that genuinely displays MORE is resolve-successful-conditions.
By default a FAILING check resolves its placeholders - "[STATUS] (502) == 200" -
while a PASSING one drops the value and shows only "[STATUS] == 200". With it
on, a healthy DNS check now reads:
[DNS_RCODE] (NOERROR) == NOERROR
[BODY] (64.226.70.190) == 64.226.70.190
which says what it actually resolved to rather than merely that the assertion
held. Applied to all 27 pulled endpoints via a gatus_endpoint_default_ui that
each caller can override.
It applies to pulled endpoints ONLY: an external (push) endpoint has no `ui`
field upstream at all, because it carries no conditions - success comes from the
push. The template was initially emitting the block in both loops; emitting an
unknown key into the external-endpoints list risks a parse rejection, and a
rejected config is exactly what skip-invalid-config-update exists to survive.
Also made the page-level `ui` a pass-through dict, the same shape as
gatus_alerting, so every upstream option (description, dashboard-heading, logo,
link, favicon, buttons, custom-css, dark-mode, default-sort-by,
default-filter-by) is reachable without a variable per key. Replaces the two
one-off gatus_ui_title / gatus_ui_header variables.
Set default-sort-by: group, because the dashboard's own grouping toggle starts
OFF and remembers per browser in localStorage - without it the ten groups render
as one flat list of 86 rows for anyone who has not clicked it.
Note the limit of all this: it is configuration. Layout, card design and group
rendering come from the Vue app compiled into the binary (//go:embed static), so
changing those means forking and rebuilding the image - which would discard the
pinned-digest property the deployment relies on.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
58 lines
2.7 KiB
YAML
58 lines
2.7 KiB
YAML
---
|
|
# One invocation writes ONE file into Gatus's endpoints directory. Gatus merges
|
|
# every *.yaml under GATUS_CONFIG_PATH and appends lists, so each caller owns
|
|
# its own file and they compose without coordinating - the same shape as
|
|
# caddy_site, where each service contributes its own vhost.
|
|
|
|
# Filename stem: <name>.yaml
|
|
gatus_endpoint_name: ""
|
|
|
|
# PULLED endpoints - Gatus makes the request and evaluates conditions.
|
|
# - {name, group, url, interval, conditions: [...], alerts: [...]}
|
|
#
|
|
# A DNS check adds `dns: {query-type, query-name}` - and note that for those,
|
|
# `url` is the RESOLVER to ask, not the name being looked up.
|
|
# A domain-expiry check is just `url: <domain>` with a [DOMAIN_EXPIRATION]
|
|
# condition; it uses WHOIS/RDAP and needs no scheme.
|
|
gatus_endpoint_pulled: []
|
|
|
|
# EXTERNAL endpoints - the host pushes its own result. Gatus never reaches out,
|
|
# which is what makes this work for machines behind NAT and for state that has
|
|
# no pollable surface at all (disk usage, ZFS health, UPS mains).
|
|
#
|
|
# - {name, group, token, heartbeat, alerts: [...]}
|
|
#
|
|
# `heartbeat` is the important one: if nothing reports within that window Gatus
|
|
# alerts. That is what makes a push check detect its own failure - a dead timer
|
|
# looks exactly like a dead host, which is the correct reading.
|
|
gatus_endpoint_external: []
|
|
|
|
# Where the files live. Matches roles/gatus.
|
|
gatus_config_dir: /opt/gatus/config
|
|
gatus_endpoints_dir: "{{ gatus_config_dir }}/endpoints"
|
|
gatus_gid: 10001
|
|
|
|
# Alerts attached to every endpoint in this file that does not specify its own.
|
|
#
|
|
# Gatus's provider-level `default-alert` only supplies DEFAULTS - an endpoint
|
|
# still has to opt in with `alerts: - type: signal` or it alerts on nothing at
|
|
# all. With ~90 endpoints that cannot be written by hand, so it is applied here.
|
|
#
|
|
# failure-threshold is set by the CALLER, because the right value depends on the
|
|
# check's cadence and there is no single correct default. See the note in
|
|
# infra/400_host_monitoring.yml.
|
|
gatus_endpoint_default_alerts: []
|
|
|
|
# UI options applied to every PULLED endpoint that does not set its own.
|
|
#
|
|
# Only pulled endpoints can carry this - an external (push) endpoint has no `ui`
|
|
# field at all, because it has no conditions to display.
|
|
#
|
|
# Note that every hide-* option already defaults to false upstream, so there is
|
|
# nothing to un-hide. The one setting that genuinely shows MORE is
|
|
# resolve-successful-conditions: by default a failing check displays the real
|
|
# value - "[STATUS] (502) == 200" - while a passing one drops it and shows only
|
|
# "[STATUS] == 200". Turning it on resolves both, so a healthy DNS check shows
|
|
# the IP it actually resolved rather than just the assertion.
|
|
gatus_endpoint_default_ui:
|
|
resolve-successful-conditions: true
|