Topic
Flux
Category
Topics
2 issues matching filters
Flux
- Field NotesJun 29, 2026
The Grafana dashboard built for the floor, not the engineer, and the reader retrofit that finally closed
Nine issues built a stack that ingests on a bounded disk, scores with a gated detector, and keeps a long-term feature record. None of it reaches the maintenance team until the page a technician opens at 6 a.m. answers the one question that brought them to it. Issue 10 builds that page: a single-asset status view backed by the feature bucket from issue 09, anomaly annotations driven off the same flags that fire the alert, and a data link that jumps from a flagged point to its pinned raw window. Plus the 30-day reader retrofit, now closed.
The stack the series has built across nine issues now ingests on a bounded disk, scores vibration with a gated detector, and keeps a long-term per-window feature record, but none of that work reaches the maintenance team, because the only human-facing output so far is a single Pushover alert that announces a problem without locating it. Issue 10 builds the Grafana dashboard the series has referenced since issue 05, and the governing decision is that it is built for the floor technician and not for the engineer who built the pipeline. The two read the same data with different questions. The engineer wants the detector's score distribution and the false-positive rate; the technician, paged at 6 a.m., wants to know which asset, how serious, since when, and what to look at. The layout answers the technician's sequence in order: a per-asset status row at the top that is green or red and nothing in between, the feature trends from the issue 09 feature bucket below it so the degradation is visible as a line rather than a single threshold crossing, and anomaly annotations drawn from the same flags that fire the alert so the page and the page that paged them agree. A Grafana data link on each flagged point carries the technician to the pinned raw window from issue 09, one click from the alert to the waveform. The high-rate vibration channel is never drawn at full rate, because a two-kilohertz series renders as a smear and tells the floor nothing; the dashboard shows features and downsampled trends and reserves the raw samples for the forensic drill-down. The whole dashboard is provisioned from a JSON model checked into git rather than assembled by hand in the browser, so it is reproducible and survives a VM rebuild. The recurring bill is unchanged at $5.50 a month. And the reader retrofit case study that has run open since issue 03 reached day 30 this week and is reported in full, with the reader's photographs and the comparison against the plant's existing monitoring contract.
Grafana·Dashboards·Influxdb·Flux·Observability·Human Factors - Field NotesJun 22, 2026
Retention and downsampling on InfluxDB OSS, and why the detector cannot eat a downsampled stream
The test cell's bucket is filling, and the question the series deferred five times is forced. A year of full-rate vibration does not fit on a 40 GB disk, so the raw stream gets a short window and the long-term record gets built from rollups. The catch is that the anomaly detector from issues 06 through 08 reads the exact transient structure that downsampling throws away, so the long-term memory has to be the feature table, not a coarsened waveform.
Since issue 05 the test cell has written full-rate vibration and five process variables into a single InfluxDB OSS bucket on the 40 GB Hetzner CX22, and every issue since has deferred the question of what happens when the disk fills. Issue 09 forces it, because the bucket is filling on a measurable schedule. Two assets streaming an accelerometer channel at two kilohertz plus their process variables produce on the order of a gigabyte a day after compression, which fills the usable disk in roughly three weeks, so retention is no longer optional. The naive fix, a short retention period on the raw bucket, drops history the plant will want the day a bearing fails. The standard fix, downsampling raw data into coarser rollup buckets, is lossy in exactly the dimension that matters, because the anomaly detector from issues 06 through 08 computes its features from the transient shape of the waveform and a one-minute mean keeps none of it. The resolution is to separate the two records the system actually needs. The raw waveform is a short forensic buffer with a seven-day retention period. The long-term memory is the per-window feature vector the edge already computes, which is small enough to keep at full event cadence for years. Downsampled rollups serve the dashboards. A flagged anomaly triggers a pin that copies its raw window into a permanent forensic bucket before the retention sweep can delete it, because a retention policy is a destructive operation on a timer and the window you needed is always the one that just expired. The cost line is unchanged at $5.50 a month plus an optional storage box for cold export, and the reader retrofit window has still not closed.
Influxdb·Retention·Downsampling·Flux·Telegraf·Time Series