perf(search): parallelize resource searches with errgroup

[DioCrafts]

perf(search): Parallelize resource searches with errgroup & remove background refresh

Search downloads full resource lists sequentially

Problem

SearchHandler.Search() iterated over up to 9 resource types sequentially. Each call to a search function triggered a full K8sClient.List() + name filtering round-trip. With 9 resource types, the worst-case latency was O(N × avg_latency) — roughly 9× the cost of a single resource search.

Additionally, GlobalSearch() on a cache hit launched a fire-and-forget background goroutine (go func() { h.Search(copiedCtx, ...) }()) to refresh the cache. This caused:

• Unbounded goroutine spawning: Every cached search hit spawned a new goroutine
• Wasted API server bandwidth: 9 sequential List calls per background refresh
• Race potential: gin.Context.Copy() used to share state with background goroutine
• Negligible UX benefit: With a 10-minute TTL, stale-while-revalidate provides minimal freshness improvement

Solution 1 — Parallel search with errgroup

Replaced the sequential for name, searchFunc := range searchFuncs loop with errgroup.WithContext() parallel execution:

resultSlices := make([][]common.SearchResult, len(entries))
g, _ := errgroup.WithContext(context.Background())

for i, entry := range entries {
    g.Go(func() error {
        results, err := entry.fn(c, q, int64(limit))
        if err != nil {
            log.Printf("search: resource %q failed: %v", entry.name, err)
            return nil // non-fatal
        }
        resultSlices[i] = results
        return nil
    })
}
_ = g.Wait()

Key design decisions:

• Each goroutine writes to its own pre-allocated index in resultSlices — no mutex needed
• Errors are logged and skipped (non-fatal) — one failing resource type doesn't break the entire search
• Results are merged, sorted, and truncated after all goroutines complete
• gin.Context is read-safe for concurrent access (only MustGet("cluster") is called)

Latency reduction: O(N × avg) → O(max) = ~9× improvement in worst case

Solution 2 — Remove background refresh goroutine

Removed the stale-while-revalidate pattern from GlobalSearch():

// BEFORE (removed):
copiedCtx := c.Copy()
go func() {
    _, _ = h.Search(copiedCtx, query, limit)
}()

// AFTER: just serve cached results
c.JSON(http.StatusOK, response)
return

Cache entries now expire naturally after the 10-minute TTL. The next request after expiry triggers a fresh (now parallel!) search.

Dead code removed

• sync.Mutex import and variable (not needed with index-based writes)
• gin.Context.Copy() call (no longer needed without background goroutine)

Tests added (4 new, 8 total — all PASS)

Test	Validates
TestSearchParallelExecution	Goroutines run concurrently (timing assertion + atomic max-concurrency counter)
TestSearchPartialFailure	One failing resource type doesn't break others
TestGlobalSearchCacheDoesNotTriggerBackgroundRefresh	Cache hit does NOT spawn background goroutine (atomic call counter)
TestSearchEmptyResourceFuncs	Graceful handling of zero searchable resource types

Files changed

File	Changes
pkg/handlers/search_handler.go	Parallel errgroup search, removed bg refresh, dead code cleanup
pkg/handlers/search_handler_test.go	4 new tests
go.mod	golang.org/x/sync promoted from indirect to direct dependency

Performance impact

Metric	Before	After
Search latency (9 types)	~9 × avg API call	~1 × max API call
Background goroutines per cached hit	1 (unbounded)	0
API calls per cached hit	9 (sequential bg refresh)	0
Memory per cached search	+1 gin.Context.Copy()	0

[DioCrafts]

CI fail already fixed @zxh326

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 521684d2d0

ℹ️ About Codex in GitHub

Codex has been enabled to automatically review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

When you sign up for Codex through ChatGPT, Codex can also answer questions or update the PR, like "@codex address that feedback".

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 86da2bb8c1

ℹ️ About Codex in GitHub

Codex has been enabled to automatically review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

When you sign up for Codex through ChatGPT, Codex can also answer questions or update the PR, like "@codex address that feedback".

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 17d1ff3801

ℹ️ About Codex in GitHub

Codex has been enabled to automatically review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

When you sign up for Codex through ChatGPT, Codex can also answer questions or update the PR, like "@codex address that feedback".

[DioCrafts]

fix lint on test file