Demo preview ConnectedA
TRAFFIC OPERATIONS / V2
Traffic overview.
Activity and health of your simulation.
Rates use the last complete minute across all workers.
Target vs observed QPS
Effective candidate target versus last complete minute. Daily targets use the remaining shared budget, allocated in proportion to channel caps; short-term allocation can vary. A recently changed target may differ from that minute.
| TAG / CHANNEL | CONFIGURED CAP | EFFECTIVE TARGET | CANDIDATES | ACHIEVED % | FILTERED / S | FILTERED % | PRIMARY VAST QPS | AUCTIONS | VAST HTTP | TRACKING HTTP | MEDIA HTTP |
|---|
Request activity
Requests per minute
Requests Errors
Waiting for data…
Health
Worker —Resource metrics describe the serving PM2 worker. Counters are shared across this installation.
Statistics by tag
Lifetime totals for this installation. Impressions count playback sessions since the counting update.
| TAG / CHANNEL | REQUESTS | INLINE RECEIVED | INLINE ACCEPTED | MEDIA VALIDATED | PLAYBACK STARTED | COMPLETED ADS | INLINE REJECTED | INLINE SKIPPED | PLAYBACK FAILED | PLAYBACK STOPPED | IMPRESSIONS | NO FILL | ERRORS | PIXEL ERRORS | COMPLETE PIXELS |
|---|
Report filters
Last 60 days · combined across all workers on this installation
Filters use exact matches. Choose values from each menu; All includes every value. Remove an individual selection with its x button. Missing data is shown as “—”. Proxy / HTTP status filters apply to HTTP diagnostics; reason applies to diagnostic tables. Operation applies to HTTP and DB. Traffic totals stay independent of these diagnostic filters.
Live QPS by tag
Waiting for the last complete minute.
All workers. Rates per second, independent of the report date and filters below. Tags shared by multiple channels are combined. The first minute after startup may be partial.
| Tag | Configured candidate QPS | Candidate QPS | Filtered / s | Filtered % | Accepted / s | Auctions / s | Primary VAST QPS | All VAST HTTP QPS | No fill / s | Errors / s | Inline received / s | Impressions / s | Tracking HTTP QPS | Media HTTP QPS | Media downloaded / s | Media failed / s |
|---|
Primary VAST counts first HTTP request starts to the configured platform. All VAST includes Wrapper requests. Rates use the last complete minute; the platform may use a different averaging window. New transport counters show — until measured; HTTP redirects inside a request are not separate counts.
Traffic insights
Last complete minutes, all workers and all tags. Independent of historical report filters. Channel counters and VAST latency samples start with this release.
Green: candidates / s. Blue dashed: current configured target (not historical settings). Orange: no fill / s. Red: errors / s.
Processing funnel
Event counts in this window, not a cohort. Multiple ads, wrappers and delayed events can produce ratios above 100%. No fill is a separate response outcome.
| Stage | Count | Ratio to preceding stage |
|---|
Channels and QPS diagnosis
Last complete minute. A runtime state is evidence, not a definitive cause of a previous-minute shortfall. No HTTP or DB cause is inferred without evidence.
| Channel / table / tags | Configured candidate QPS | Candidate QPS | Filtered / s | Filtered % | Accepted / s | Primary VAST QPS | All VAST HTTP QPS | Runtime / diagnosis |
|---|
Primary VAST = first HTTP request starts to the configured platform; All VAST includes Wrapper requests. Filtered candidates never reach the platform. New counters display — until measured. The calculator estimates candidate QPS from the observed pass rate and never saves configuration. Check Filtering with the same tag for the reasons; multiple reasons can match one candidate.
Last 5 minutes vs previous 5
| Metric | Previous | Current | Change |
|---|
Traffic statistics
Choose a date range and run the report.
Columns
PostgreSQL queries
For the selected period and DB table. Queries are not attributed to tags, applications or countries.
| TABLE / OPERATION | DB REQUESTS | SUCCESSFUL | ERRORS | ROWS | AVG MS | P50 MS ≈ | P95 MS ≈ | P99 MS ≈ |
|---|
Filtering reasons
Summarized by reason across matching tags, channels and proxies. No IP breakdown. A record can match several reasons; their sum is not the number of uniquely filtered records.
| REASON | MATCHES | % OF SELECTED CANDIDATES |
|---|
HTTP latency and bandwidth
All workers · selected period · approximate histogram percentiles, including failures. Duration includes response body transfer; local rejections measure queue wait. Received payload bytes include partial streams and exclude headers, TLS, redirect bodies and outbound traffic. VAST/media bytes are after decompression; tracking bytes retain content encoding. These are not billable wire bytes.
| TYPE | ATTEMPTS | ERRORS | ERROR % | RECEIVED BYTES | PROXY RECEIVED BYTES | P50 MS ≈ | P95 MS ≈ | P99 MS ≈ |
|---|
HTTP status and error breakdown
Transport attempts include local overload rejections (status 0). Error % is within each row. No URLs, credentials or raw error messages are stored.
| TAG | CHANNEL | TYPE | PROXY | HTTP STATUS | REASON | ATTEMPTS | ERRORS | ERROR % | RECEIVED BYTES |
|---|
Ad outcomes and failure reasons
Failure events can exceed auctions because one auction may contain multiple ads and download attempts.
| TAG | CHANNEL | REASON | Description | FAILURE EVENTS | INLINE REJECTED | PLAYBACK FAILED | INLINE SKIPPED | PLAYBACK STOPPED |
|---|
Verification providers
Observed VAST providers and HTTP outcomes by tag, source channel, subid and table. Selected ads count once per provider across the wrapper chain. Resources are declared verification scripts. HTTP OK means delivery succeeded, not a passed fraud or viewability check. OMID requests do not prove script execution. Not executed counts attempted verificationNotExecuted notifications. Unknown providers are grouped without recording URLs or vendor strings. Collection starts with this release; prior traffic cannot be reconstructed.
| Tag | Channel ID | Subid | DB table | Provider | Observation | Selected ads | Resources | HTTP attempts | HTTP OK | HTTP errors | Not executed notifications |
|---|
How metrics are calculated
Ad stages: Inline received counts InLine ads in successfully parsed responses, including invalid ads and unused backups. Inline accepted is the existing Inline counter: ads selected for media processing. Media validated counts one successfully checked media candidate per ad, not download attempts. Playback started counts actual OMID start or simulated start once; Completed ads counts completion once. Inline rejected counts invalid/policy-rejected ads; Inline skipped counts unused backups and ads skipped by the work limit. Playback failed counts an accepted ad failing after retries; Playback stopped counts intentional early simulation stops. Outcome reasons are shown below. New stage counters start with this release; previous history is not backfilled. Malformed XML and responses rejected before XML parsing cannot supply a reliable Inline received count. In-flight ads and interrupted processes can leave stages incomplete. HLS validation covers the manifest, not its segments.
Multiple ads: one response may contain several Inline ads. Completed ads count playbacks reaching completion once; Complete pixels count successful completion URLs. Priced impressions include zero USD CPM; Missing price counts impressions without supported pricing. Priced CPM excludes missing prices and legacy records without price classification. Detailed traffic outcomes use the auction start hour; overview QPS uses actual event times. Earlier history retains its original attribution.
Candidates: records from the SQL cache that reached validation. Filtered: rejected candidates, counted once. Accepted: candidates that passed validation. Auctions: VAST chains started. Requests: individual VAST HTTP steps, including wrappers. Inline ads: accepted InLine ads, including multiple ads per response. No fill counts empty VAST responses separately from technical errors after this update; earlier error totals retain their original classification. VAST 303 remains in diagnostics. Error percentage counts a response once even when several ads fail. Filled auctions count auctions with at least one accepted ad. Ads / auction = Inline ads / auctions. Fill rate = filled auctions / auctions. Coverage = successful impressions / Inline.
An HTTP OK impression means at least one successful impression endpoint in a playback session; multiple endpoints do not increase the impression count. An error means all attempts failed. External validity and viewability are not measured. Revenue uses the first declared VAST Pricing when it is a valid USD CPM. When Pricing is absent, a single unambiguous adrta.com pricePaid value is used as USD CPM, as confirmed by the platform owner. Each successful impression contributes CPM / 1000 once; failed impressions contribute nothing. Missing or unsupported prices contribute no revenue. A dash means impressions exist but none have a known price or recorded revenue. A known zero CPM is displayed as zero. If some prices are missing, revenue covers only the recorded amounts. It is an estimate, not a confirmed payment. pricePaid (adrta.com) and wp (e.adstir.com) are independent averages of observed impression URL values per selected Inline, including failed playbacks. pricePaid is confirmed USD CPM and may supply missing Pricing for new successful impressions. wp currency and units remain unverified and wp never contributes to revenue. Neither observed value changes macro replacement. Repeated identical values count once per ad; conflicting values are omitted. Missing values display a dash. Collection starts with the observed-price release. Revenue fallback starts with the fallback release; older observed averages cannot be backfilled into revenue because they include failed playbacks.
Media downloaded: completed transfers; download failures and invalid media are separate counters. Downloads cancelled and Validation cancelled identify interruptions by the auction deadline. Large reports may prepare daily summaries in several bounded steps; partial totals are never displayed. A downloaded file may still be invalid. Retries count as separate attempts. Media bytes count completed transfers. DB queries: source-data SQL calls, including failed attempts. Historical statistics writes and report queries use a separate PostgreSQL connection and are excluded to avoid recursive counting. A cache hit is not a SQL query. Rows: returned by SELECT or actually affected by INSERT. Duration includes connection pool wait. Basic historical metrics and DB record reuse are retained for 60 days. Application name, bundle, website and store URL breakdowns follow the configured detail retention (7 days on new installations). Older periods cannot use these detailed groupings or filters; the report shows a clear error rather than partial totals. Day and month groups are calculated in the selected time zone; monthly results cover only the selected dates. History already expired before this update cannot be restored. The current hour may change while being read. Missing or unverified dimensions are not inferred.
Daily unique records
Shared across all workers and channels using the source. The cycle ends at 00:00 UTC. Exhaustion stops the source; there is no same-day wraparound. New rows join the next prepared cycle.
| METRIC | VALUE |
|---|
Filter actions today
Reason totals may overlap. Queue states count unique IDs. Restoring an exclusion makes the row eligible for the next cycle; deleted rows cannot be restored.
| REASON / STATE | ACTION | COUNT |
|---|
Simulation channels
QPS is the candidate rate per tag across all workers, before filtering. Capacity and DB delays can lower the achieved rate. Changes apply immediately.
| CHANNEL / TABLE | RUNTIME | QPS / TAG | TAG IDS | PLATFORM | LAST CANDIDATE | LAST ERROR | CONTROL |
|---|
Platforms
URL templates for configured channels.
| KEY | URL TEMPLATE |
|---|
Remote DB profiles
Loading profiles…
Selects the source for remote tables and future downloads. Local tables, channels and QPS stay unchanged. Finish queued transfers before switching. Deployment default preserves the GitLab connection.
New profile
Remote tables
Analyze IDFV runs one read-only scan at a time. Large tables can take several minutes. Values are trimmed and case-folded; empty/null values count as missing. Results are cached snapshots; run again after data changes. Duplicate rows = extra rows beyond one per nonempty IDFV. No IDFV values are downloaded.
| TABLE | SIZE | ESTIMATED ROWS | UNIQUE IDFV | REPEATED IDFV | DUPLICATE ROWS | MISSING IDFV | ANALYSIS | ACTIONS |
|---|
Elapsed: Remaining:
Local tables & archives
Archive creates and verifies a compressed local backup, then frees data and index pages while preserving the table identity and schema. Detach every channel (including paused channels and IDFV dependencies) first. Unarchive restores original rows and IDs without restarting channels. Local backups share the server’s disk; savings depend on compression. Imports, archives and restores share one background job slot.
| TABLE | STATUS | ARCHIVED ROWS | BACKUP SIZE / DATE | ACTION |
|---|
Transfer history
| TABLE | STATUS | STARTED | ELAPSED | VALIDATION | ERROR |
|---|
Proxies
Proxy list for this host: · 0 entries. VAST and media downloads stay direct.
| HOST / IP | PORT | USERNAME | PASSWORD | HEALTH | LAST CHECK |
|---|