Guard report_shape_db_stats against :noproc during deployments

[alco]

Summary

Electric.StackSupervisor.Telemetry.report_shape_db_stats/2 calls ShapeDb.statistics/1, which resolves the GenServer pid and then issues a GenServer.call. During stack restarts the underlying server can exit between those two steps, producing an uncaught :exit, :noproc from the telemetry poller and flooding Sentry with errors on every deployment.

The sibling function report_retained_wal_size already handles this race. This change mirrors its try/catch :exit, {:noproc, _} pattern. Surgical, no behavioural change on the happy path.

Refs electric-sql/stratovolt#1452.

Test plan

• mix compile --warnings-as-errors is clean
• mix test test/electric/stack_supervisor_test.exs passes

[claude]

Claude Code Review

Summary

This PR guards report_shape_db_stats/2 against a TOCTOU race where the ShapeDb.Statistics GenServer can exit between GenServer.whereis/1 and GenServer.call/2, causing uncaught :exit, :noproc exceptions that flood Sentry during deployments. The fix mirrors the existing pattern in the sibling report_retained_wal_size/3. The change is correct and surgical.

What's Working Well

• The root cause diagnosis is accurate: Statistics.current/1 resolves the pid then calls it, creating a window for :noproc (packages/sync-service/lib/electric/shape_cache/shape_status/shape_db/statistics.ex:19-27).
• The fix correctly mirrors report_retained_wal_size/3, using the same try/catch :exit, {:noproc, _} pattern (telemetry.ex:106-108).
• report_write_buffer_size doesn't need the same treatment — pending_operations_count uses a direct ETS lookup with its own rescue ArgumentError (write_buffer.ex:157-161), not a GenServer call.
• Changeset file is included and correctly scoped to @core/sync-service.

Issues Found

Suggestions (Nice to Have)

Inconsistent error containment vs. report_retained_wal_size

File: packages/sync-service/lib/electric/stack_supervisor/telemetry.ex:153-156

report_retained_wal_size/3 catches both :exit, {:noproc, _} and a broad type, reason -> fallback that logs a warning and returns :ok. report_shape_db_stats/2 only catches the :noproc case. A GenServer call timeout (:exit, {:timeout, {GenServer, :call, _}}) or an unexpected raise from OpenTelemetry.execute would still propagate out of the try block uncaught.

For a telemetry poller callback, silently swallowing (or logging) all errors is typically the right call — a reporting failure should never crash the poller. The existing pre-PR code also had this gap, so the PR doesn't regress anything, but it's worth making the two functions consistent:

catch
  :exit, {:noproc, _} ->
    :ok

  type, reason ->
    Logger.warning(
      "Failed to query shape DB stats\nError: #{Exception.format(type, reason)}",
      stack_id: stack_id
    )
end

Missing @spec annotation

report_retained_wal_size/3 has @spec report_retained_wal_size(Electric.stack_id(), binary(), map()) :: :ok. report_shape_db_stats/2 has none. Minor, but worth adding for consistency:

@spec report_shape_db_stats(Electric.stack_id(), map()) :: :ok

Issue Conformance

The PR references electric-sql/stratovolt#1452 (a private repo), so the linked issue isn't publicly visible. Based on the PR description, the fix is clearly scoped: match the :noproc guard pattern that already exists in report_retained_wal_size. No scope creep or missing pieces visible from the public side.

---

Review iteration: 1 | 2026-04-21

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 89.24%. Comparing base (365dd17) to head (a6819d5).
⚠️ Report is 3 commits behind head on main.
✅ All tests successful. No failed tests found.

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #4142      +/-   ##
==========================================
+ Coverage   89.20%   89.24%   +0.03%     
==========================================
  Files          25       25              
  Lines        2520     2520              
  Branches      636      635       -1     
==========================================
+ Hits         2248     2249       +1     
+ Misses        270      269       -1     
  Partials        2        2              

Flag	Coverage Δ
packages/experimental	87.73% <ø> (ø)
packages/react-hooks	86.48% <ø> (ø)
packages/start	82.83% <ø> (ø)
packages/typescript-client	94.36% <ø> (+0.05%)	⬆️
packages/y-electric	56.05% <ø> (ø)
typescript	89.24% <ø> (+0.03%)	⬆️
unit-tests	89.24% <ø> (+0.03%)	⬆️

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.