We know how many Stacker licenses we've sold but not how many copies were downloaded. Today the only trace is the StackerDownload Seq log line per HTTP response (DownloadMeter.cs): no IP, geo, user agent, referrer, user or website-vs-updater split, no durable store, nothing on any dashboard.
Plan
Backend (Domain/ReleaseFeeds): DownloadMeter captures IP, UA, referrer, src/utm_*, optional download ticket / bearer user, then queues it. A hosted writer adds GeoIP and a visitor hash and stores it in release-downloads (kept) plus release-download-addresses (raw IP, 90-day TTL). Admin endpoints: api/stacker/downloads/admin/{analytics,recent}, a stackerDownloads section on api/admin/dashboard, and a Seq backfill (counts only, since 2026-09-11).
Stacker app: the updater sends User-Agent: SpikerSoft-Stacker/<ver> (<os>; <arch>) and ?src=updater.
Angular: /stacker links carry src/utm/ticket; a staff-dashboard tile; an /admin/stacker-downloads page with geo map and live table; download tiles on the /finance/sales Stacker tab; privacy copy update.
Statusboard: a new Downloads page (KPIs, per-version and per-country bars, live feed).
Decisions: raw IP kept 90 days; hashed visitor + geo kept; statusboard stays on the TX2; counts-only backfill from Seq.
We know how many Stacker licenses we've sold but not how many copies were downloaded. Today the only trace is the `StackerDownload` Seq log line per HTTP response (`DownloadMeter.cs`): no IP, geo, user agent, referrer, user or website-vs-updater split, no durable store, nothing on any dashboard.
## Plan
- **Backend** (`Domain/ReleaseFeeds`): `DownloadMeter` captures IP, UA, referrer, `src`/`utm_*`, optional download ticket / bearer user, then queues it. A hosted writer adds GeoIP and a visitor hash and stores it in `release-downloads` (kept) plus `release-download-addresses` (raw IP, 90-day TTL). Admin endpoints: `api/stacker/downloads/admin/{analytics,recent}`, a `stackerDownloads` section on `api/admin/dashboard`, and a Seq backfill (counts only, since 2026-09-11).
- **Stacker app**: the updater sends `User-Agent: SpikerSoft-Stacker/<ver> (<os>; <arch>)` and `?src=updater`.
- **Angular**: `/stacker` links carry `src`/utm/ticket; a staff-dashboard tile; an `/admin/stacker-downloads` page with geo map and live table; download tiles on the `/finance/sales` Stacker tab; privacy copy update.
- **Statusboard**: a new Downloads page (KPIs, per-version and per-country bars, live feed).
Decisions: raw IP kept 90 days; hashed visitor + geo kept; statusboard stays on the TX2; counts-only backfill from Seq.
"Statusboard stays on the TX2" no longer holds. The wall display (statusboard tui on tty1) now runs on xavier-2; the monitor moved from dreamstream2 and has been live since ~05:50Z. The TX2s are being drained and retired with GlusterFS.
The collector is unchanged: it still runs on the x86 manager (statusboard stack) on :9300.
Deploying the Downloads page (0.2.0): after spikerj/spikersoft-infrastructure#631 merges and the collector is deployed, re-run the display installer on xavier-2 so the TUI binary is 0.2.0: cd /mnt/infrastructure && sudo git pull && sudo bash statusboard/host/install-display.sh
Other moves the backend/Angular halves touch, none needing code changes:
Seq is on xavier-6. Same URL, with history before and after the move intact, so the StackerDownload backfill (--until = deploy time) reads the same store.
Keycloak is on xavier-7. The same realm, clients and roles were copied from its database, so the spikersoft-statusboard client-credentials login and the staff/finance policies behave the same.
RabbitMQ is on xavier-5, and OpenBao on xavier-1/3/4. Analytics:VisitorHashKey and the ticket key read from Bao as before.
CI deploys now run from xavier-1's local /mnt/infrastructure. Workflows are unchanged.
MongoDB stays on SERVER, so release-downloads and the TTL collection are unaffected.
Redis moves next, one node at a time with its data, and the cluster stays up. The dashboard aggregate cache will not be flushed.
**Update: the TX2 → Xavier move changes one decision here** (spikerj/spikersoft-infrastructure#621, Phase 1 of the GlusterFS retirement, 2026-10-13).
- **"Statusboard stays on the TX2" no longer holds.** The wall display (`statusboard tui` on tty1) now runs on **xavier-2**; the monitor moved from dreamstream2 and has been live since ~05:50Z. The TX2s are being drained and retired with GlusterFS.
- The collector is unchanged: it still runs on the x86 manager (`statusboard` stack) on `:9300`.
- The security note in spikerj/spikersoft-infrastructure#631 still applies: the snapshot carries download IPs, cities and ISPs, so it stays LAN-only.
- **Deploying the Downloads page (0.2.0):** after spikerj/spikersoft-infrastructure#631 merges and the collector is deployed, re-run the display installer on xavier-2 so the TUI binary is 0.2.0:
`cd /mnt/infrastructure && sudo git pull && sudo bash statusboard/host/install-display.sh`
- **Other moves the backend/Angular halves touch, none needing code changes:**
- **Seq** is on xavier-6. Same URL, with history before and after the move intact, so the `StackerDownload` backfill (`--until` = deploy time) reads the same store.
- **Keycloak** is on xavier-7. The same realm, clients and roles were copied from its database, so the `spikersoft-statusboard` client-credentials login and the staff/finance policies behave the same.
- **RabbitMQ** is on xavier-5, and **OpenBao** on xavier-1/3/4. `Analytics:VisitorHashKey` and the ticket key read from Bao as before.
- **CI deploys** now run from xavier-1's local `/mnt/infrastructure`. Workflows are unchanged.
- **MongoDB** stays on SERVER, so `release-downloads` and the TTL collection are unaffected.
- **Redis** moves next, one node at a time with its data, and the cluster stays up. The dashboard aggregate cache will not be flushed.
Audit 2026-10-14. All four halves are merged and live:
part
PR
verified
Backend store + endpoints
spikersoft-backend#1388
deployed; endpoints answer 401 anonymously; rows being written since 2026-10-13T16:55Z
Stacker updater UA / src=updater
spikersoft-stacker#975
in v2026.10.13f+; updater shows as a source in live data → closed spikersoft-stacker#976
Angular page, tile, finance, tagged links
spikersoft-angular#1587
in the production bundle → closed spikersoft-angular#1588
Statusboard Downloads page
spikersoft-infrastructure#631
collector 0.2.3 serving downloads; wall-display binary on xavier-2 not checked
Left before this epic closes:
Seq backfill not run (no seq-backfill rows; all-time = 7). Window: --since 2026-09-11T00:00:00Z --until 2026-10-13T16:55:59Z. Tracked on spikersoft-backend#1390.
Confirm the xavier-2 wall shows page 8 (spikersoft-infrastructure#634).
**Audit 2026-10-14.** All four halves are merged and live:
| part | PR | verified |
|---|---|---|
| Backend store + endpoints | spikersoft-backend#1388 | deployed; endpoints answer 401 anonymously; rows being written since 2026-10-13T16:55Z |
| Stacker updater UA / `src=updater` | spikersoft-stacker#975 | in v2026.10.13f+; `updater` shows as a source in live data → **closed** spikersoft-stacker#976 |
| Angular page, tile, finance, tagged links | spikersoft-angular#1587 | in the production bundle → **closed** spikersoft-angular#1588 |
| Statusboard Downloads page | spikersoft-infrastructure#631 | collector 0.2.3 serving `downloads`; wall-display binary on xavier-2 not checked |
Left before this epic closes:
1. **Seq backfill not run** (no `seq-backfill` rows; all-time = 7). Window: `--since 2026-09-11T00:00:00Z --until 2026-10-13T16:55:59Z`. Tracked on spikersoft-backend#1390.
2. Confirm the xavier-2 wall shows page 8 (spikersoft-infrastructure#634).
Blocking a user prevents them from interacting with repositories, such as opening or commenting on pull requests or issues. Learn more about blocking a user.
We know how many Stacker licenses we've sold but not how many copies were downloaded. Today the only trace is the
StackerDownloadSeq log line per HTTP response (DownloadMeter.cs): no IP, geo, user agent, referrer, user or website-vs-updater split, no durable store, nothing on any dashboard.Plan
Domain/ReleaseFeeds):DownloadMetercaptures IP, UA, referrer,src/utm_*, optional download ticket / bearer user, then queues it. A hosted writer adds GeoIP and a visitor hash and stores it inrelease-downloads(kept) plusrelease-download-addresses(raw IP, 90-day TTL). Admin endpoints:api/stacker/downloads/admin/{analytics,recent}, astackerDownloadssection onapi/admin/dashboard, and a Seq backfill (counts only, since 2026-09-11).User-Agent: SpikerSoft-Stacker/<ver> (<os>; <arch>)and?src=updater./stackerlinks carrysrc/utm/ticket; a staff-dashboard tile; an/admin/stacker-downloadspage with geo map and live table; download tiles on the/finance/salesStacker tab; privacy copy update.Decisions: raw IP kept 90 days; hashed visitor + geo kept; statusboard stays on the TX2; counts-only backfill from Seq.
Update: the TX2 → Xavier move changes one decision here (spikerj/spikersoft-infrastructure#621, Phase 1 of the GlusterFS retirement, 2026-10-13).
statusboard tuion tty1) now runs on xavier-2; the monitor moved from dreamstream2 and has been live since ~05:50Z. The TX2s are being drained and retired with GlusterFS.statusboardstack) on:9300.cd /mnt/infrastructure && sudo git pull && sudo bash statusboard/host/install-display.shStackerDownloadbackfill (--until= deploy time) reads the same store.spikersoft-statusboardclient-credentials login and the staff/finance policies behave the same.Analytics:VisitorHashKeyand the ticket key read from Bao as before./mnt/infrastructure. Workflows are unchanged.release-downloadsand the TTL collection are unaffected.Audit 2026-10-14. All four halves are merged and live:
src=updaterupdatershows as a source in live data → closed spikersoft-stacker#976downloads; wall-display binary on xavier-2 not checkedLeft before this epic closes:
seq-backfillrows; all-time = 7). Window:--since 2026-09-11T00:00:00Z --until 2026-10-13T16:55:59Z. Tracked on spikersoft-backend#1390.