Skip to content

local: Matomo analytics tracking (never push upstream)#4

Merged
efiten merged 1 commit into
masterfrom
local/matomo-analytics
Apr 29, 2026
Merged

local: Matomo analytics tracking (never push upstream)#4
efiten merged 1 commit into
masterfrom
local/matomo-analytics

Conversation

@efiten
Copy link
Copy Markdown
Owner

@efiten efiten commented Apr 29, 2026

What

Adds the Matomo tracking snippet (analytics.fiten.eu, site ID 8) to public/index.html just before </body>.

Important

This PR is local-only. It must be kept on efiten/master (the deploy branch) and never cherry-picked to any upstream PR targeting Kpa-clawbot/meshcore-analyzer.

When rebasing upstream changes into the deploy branch, re-apply this commit on top.

Injects Matomo snippet (analytics.fiten.eu, site ID 8) before </body>.
This commit is local-only — keep on origin/master and deploy branches,
never cherry-pick to upstream PRs.

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
@efiten efiten merged commit 76f6877 into master Apr 29, 2026
6 checks passed
Kpa-clawbot pushed a commit that referenced this pull request May 16, 2026
… — failure observability, lock granularity, single cutoff, atomic contract

Five-in-one because they all touch the same hot-startup data flow:

#3 backgroundLoadFailed flag: surfaced in /api/perf map and the typed
   PerfPacketStoreStats struct. Set true iff at least one chunk errored
   during loadBackgroundChunks. backgroundLoadComplete now only signals
   'the loader returned', backgroundLoadFailed signals data integrity.

#4 atomic ordering contract: doc comment on hashMigrationComplete /
   backgroundLoadDone / backgroundLoadFailed explaining what each gates
   and the read-order invariant (Done must be read before Failed is
   meaningful).

Kpa-clawbot#6 loadChunk merge: the previous implementation held s.mu.Lock across
   the entire merge of a chunk (hundreds of ms on large chunks);
   runtime.Gosched between chunks didn't help readers/ingest that
   arrived mid-chunk. Now the prepend is one short critical section
   and the per-index merge runs in 500-packet batches with the lock
   released+reacquired between batches plus a Gosched. Counter update
   is its own tiny final section.

Kpa-clawbot#7 single cutoff in Load(): replaced the second time.Now().UTC() call
   used to seed s.oldestLoaded with the same hotCutoffStr that fed the
   load SQL. Microsecond skew at chunk borders is gone.

#3 covered by TestHotStartup_BackgroundLoadFailureSurfacesInPerf
   (RED in the previous commit, GREEN here).
Kpa-clawbot#6/Kpa-clawbot#7 covered by existing TestHotStartup_BackgroundFillsToRetention
   and TestHotStartup_ConcurrentQueryDuringBackgroundLoad — both still
   pass. bounded_load_test fixture had to switch observations.timestamp
   from RFC3339 string to unix int (production schema) so the restored
   item-1 RFC3339 subquery query path can match.
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant