feat(dashboard): histogram rate|mean card pair#938
Merged
thinkingfish merged 19 commits intoJun 1, 2026
Conversation
selection.js moved to lib/selection/selection.js in iopsystems#932; the smoke assertion was still curling /lib/selection.js and 404'ing.
Bump metriken-query 0.10.5 → 0.10.6 for histogram_irate(m). Replaces the spec'd irate(histogram_count(m)[5m]) form, which doesn't parse: PromQL grammar disallows range vectors on function-call results. histogram_irate returns an instant vector — composes with sum(...) directly and sidesteps the grammar problem.
Bumps metriken-query 0.10.7 → 0.10.8 for histogram_sum(m). Dashboard's histogram_rate_mean helper now renames the Mean card to Mean/Total and stashes histogram_sum(<selector>) on the spec as promql_query_total. The line-chart configure path renders an inline checkbox top-right that fetches and swaps to the total series on click; tooltip series label substitutes Mean ↔ Total to match the active view. Per-chart state, no persistence; pattern mirrors scatter.js's spectrum-controls.
initWasmViewer replaces the WASM TSDB but sectionResponseCache, chartsState, heatmapDataCache, and data.js's metadata cache survived across loads. The new capture's sections short-circuited to the old file's cached responses, rendering empty. First load was unaffected because the caches start empty.
…captures" This reverts commit abbe1be.
Reloading /viewer/#/overview with no ?demo/?capture param lands on the FileUpload page but the URL still showed #/overview. Drop the hash via history.replaceState when there's no capture in the URL, so the location bar matches what's rendered.
Three things were stale across captures:
- data.js's cachedMetadata held the previous file's {minTime, maxTime},
so fresh PromQL queries against the new TSDB ran on a window outside
it and returned empty.
- The active section's cached payload (and therefore the topnav's
time-range bar) wasn't refetched, because m.route's same-path
short-circuit suppressed onmatch.
- The cachedView placeholder spins on a splash until loadSection
resolves, so the eviction needs an explicit refetch kick rather
than relying on a redraw to do it.
clearMetadataCache() + evict the visible section + loadSection(...) +
m.redraw() covers all three without touching chartsState or the rest
of sectionResponseCache (both of which spun in earlier attempts).
Place it where scatter.js's spectrum-controls sit — under the title, left edge aligned with the echarts plot grid (past the y-axis gutter). Previous top-right placement collided with the expand/select-pin icons and read as right-aligned on half-width Mean cards.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
For every percentile-subtype histogram plot in the dashboard, emit an adjacent half-width Rate + Mean card pair.
histogram_mean(<sel>)(metriken-query 0.10.5+). Robust to lock-free snapshot undercount because it's a ratio.blockio_operationsfor blockio latency/size,syscallfor syscall latency — these existing rate cards are kept verbatim, the only change is adding the adjacent Mean). Falls back tosum(histogram_irate(<sel>))(metriken-query 0.10.6+) for histogram-only families —scheduler_runqueue_latency,scheduler_offcpu,scheduler_running,tcp_packet_latency.Strictly additive: no agent / parquet / exporter / MCP changes. Existing percentile cards untouched.
A small helper
SubGroup::histogram_rate_mean(title_stem, id_stem, selector, RateSource, mean_unit)keeps the pair adjacent and declarative at call sites.Why the fallback uses
histogram_irate(not the originally spec'd form)The original spec called for
sum(irate(histogram_count(<sel>)[5m])). That doesn't parse — PromQL grammar disallows range vectors on function-call results. In metriken-query's evaluation model (one value per step from the histogram column) the[5m]window is vestigial anyway, so the cleaner contract is an instant-vector function with no window. metriken-query 0.10.6 shipshistogram_irate(m)for exactly this;sum(histogram_irate(<sel>))composes directly.Drive-by
Smoke-test
selection.jspath updated for the lib/ subdir refactor in #932 (separate commit for clean blame).Test plan
cargo test -p dashboard— 39 passcargo clippy --all-targets -- -D warnings— cleanbash tests/viewer_smoke.sh— all assertions passcargo run -p dashboard -- /tmp/dump/— verified scheduler.json + network.json carrysum(histogram_irate(...)); blockio/syscall keep their counter-backed rates; every pair card is half-width