Reliability¶
The Reliability module answers one question: is BusinessObjects working for users right now, and was it working overnight.
How it works¶
The module has two pages:
| Page | What it shows | Data |
|---|---|---|
| WebI Live | Web Intelligence processing servers and activity, refreshed every few seconds | MetaMiner server metrics and BO activity. See Web Intelligence. |
| Flight Recorder | Report availability against a service level objective (SLO), with probe runs, changes and real user activity on one time axis | Synthetic probe runs, BO Auditing, the server configuration change log and server availability |
The Flight Recorder relies on synthetic probes: a probe service opens a fixed list of watched reports on a schedule, records how long each took to open (time to render, TTR), the outcome and a screenshot, and writes a heartbeat every 60 seconds. MetaMiner reads those runs and judges each report against the SLO.
TODO: document how to install and configure the report probe service (webitester)
The Flight Recorder reads runs written by the webitester probe service. Add the installation steps, how to choose the watched reports and probe interval, and which licence includes it.
Open the Flight Recorder¶
In the BO console sidebar choose Reliability, then the Flight Recorder tab.
Image of Reliability > Flight Recorder with the verdict panel, the Error-budget burn table, the timeline with three report lanes and change markers, and the three bottom panels
../../images/web-console/bo/flight-recorder.png
Controls¶
| Control | What it does |
|---|---|
| Environment list | All environments or one environment |
| Report list | All watched reports or focus on one report (grouped by environment) |
| Window | 1 h, 3 h, 6 h, 12 h, 1 d, 3 d, 7 d, 2 wk, 1 m, 3 m, 6 m |
Times are shown in your browser's time zone; the subtitle names the zone, the number of probes and environments, and how much of the window the probes actually observed.
Panels¶
| Panel | What it shows |
|---|---|
| Verdict | A plain sentence: "Reports are opening.", how many reports are failing now, or "Health unknown, not OK" when the probe service is silent. Below it, one line per outage, degraded or slow report, silent period and the change that preceded an outage. |
| Error-budget burn | Per report, how fast it spends its error budget over 1 h, 6 h, 24 h and 28 days, with the alert class: page, ticket, spent or recovered |
| Flight recorder | The timeline: one lane per watched report with each run coloured by outcome, server starts and configuration changes as markers, and real user activity from BO Auditing. Hover for the moment; click a lane to drill in. |
| Where it breaks | The stages of opening a report (for example navigation, refresh) with the worst 95th-percentile time per stage on successful runs, the previous window for comparison, and failed runs per stage |
| Who is at fault | Non-OK runs grouped by report owner, with their failure labels |
| Probe coverage | The share of report opens in BO Auditing over 28 days that the watched reports carry, the most-opened reports marked watched or not watched, and how many probe opens BO Auditing recorded (audit echo) |
The burn rules, as shown under the panel: page when 1 h burn is at least 14.4 times the allowed rate and the last run is down, or when 1 h and 6 h are both at least 6 times; ticket when 6 h and 24 h are both at least 3 times; spent when the 28-day burn reaches 1 (needs half of the 28 days observed); recovered when a report burned in the last 6 h and now passes. Partial data, probe faults and unobserved time burn nothing.
Unknown is not up
Time when the probe service was silent is excluded from availability, never counted as up. Values the page cannot know read "unknown", never 0.
Drill-in and screenshots¶
Click a report lane to open its drill-in under the timeline. It shows the report's environment, folder and state, and a table of the last 10 runs: Started, Class, Label, Owner, BO message, TTR, Alert. Click a run to see the screenshot the probe took.
Image of a Flight Recorder drill-in for one report: the last 10 runs table with one failed run selected and its screenshot shown on the right
../../images/web-console/bo/flight-recorder-drill-in.png
Screenshots are read from the probe service's screenshot folder on the MetaMiner Server machine. The probe service keeps the newest 60 per report; older runs show "Screenshot not available".
Settings¶
These properties are read by the Web Apps service when it starts. Set them in server\config\config.json (properties block) and restart the MetaMiner Server.
| Property | Default | Effect |
|---|---|---|
web.apps.reliability.slo.availability |
0.995 |
Availability objective per report (99.5 percent) |
web.apps.reliability.slo.ttr.ms |
15000 |
Time-to-render objective in milliseconds; reports whose 95th percentile exceeds it are flagged slow |
web.apps.reliability.service.users |
the Service accounts: list | Logins excluded from real user activity |
web.apps.reliability.cache.sec |
30 |
How long one computed view is reused |
web.apps.reliability.screenshots.dir |
..\webitester\screenshots (relative to the server folder) |
Where probe screenshots are read from |
Troubleshooting¶
"Health unknown, not OK: webitester has not written a single run yet."
The probe service is not installed or has never run. Start it; the page fills in on its first run.
The verdict says the probe service has been silent
Check that the probe service runs on its host. Silent time is excluded from availability until it resumes.
WebI Live is empty
See Web Intelligence for the data it needs.