feat(nethttp-mysql): add net/http + MySQL sample for replay regression by officialasishkumar · Pull Request #224 · keploy/samples-go

officialasishkumar · 2026-04-21T15:20:43Z

Describe the changes that are made

Adds a minimal nethttp-mysql sample — standard library net/http
plus database/sql with go-sql-driver/mysql — that exercises the
full MySQL connection and command phase end-to-end:

/health calls db.Ping() (forces a real command roundtrip).
/users SELECTs from a seeded users table.
/users/add INSERTs and returns the new row.

Why this sample exists: Keploy's OSS pipeline will consume it as a
regression guard for the MySQL replay command-phase read deadline.
If the proxy ever re-introduces a zero or negative deadline, the Go
driver blocks inside db.Ping() during replay, the HTTP listener
never binds :8080, and every replayed request fails with
connection reset by peer — making the regression loud in CI
rather than silently hanging on the user's side.

The companion PR in keploy/keploy (fix: prevent zero/overflow read
deadline in MySQL replay command phase) adds the pipeline script that
drives this sample and the fix for the proxy math.

Links & References

Closes: keploy/keploy#4102

🔗 Related PRs

keploy/keploy: companion fix + pipeline wiring (same author)

🐞 Related Issues

NA

📄 Related Documents

NA

What type of PR is this? (check all applicable)

Added e2e test pipeline?

👍 yes — via the companion PR in keploy/keploy which adds a
golang_docker matrix entry nethttp-mysql and the driver script
at .github/workflows/test_workflow_scripts/golang/nethttp_mysql/golang-docker.sh.

Added comments for hard-to-understand areas?

👍 yes — the package-level doc comment in main.go explains
why the app is shaped this way (regression guard).

Added to documentation?

📜 README.md — new sample-level README with local-run and
keploy record/test instructions.

Are there any sample code or steps to test the changes?

👍 yes

Local smoke test:

cd nethttp-mysql
docker network inspect keploy-network >/dev/null 2>&1 || docker network create keploy-network
docker run -d --name mysql --network keploy-network -e MYSQL_ROOT_PASSWORD=password -e MYSQL_DATABASE=testdb mysql:8.0
docker build -t nethttp-mysql:1.0 .
docker run --rm -p 8080:8080 --network keploy-network \
  -e MYSQL_HOST=mysql -e MYSQL_USER=root -e MYSQL_PASSWORD=password -e MYSQL_DATABASE=testdb \
  nethttp-mysql:1.0
curl -s localhost:8080/health
curl -s localhost:8080/users
curl -s "localhost:8080/users/add?name=charlie&email=charlie@test.com"

Self Review done?

✅ yes

Any relevant screenshots, recordings or logs?

NA

🧠 PR Semantics

PR Title: feat(nethttp-mysql): add net/http + MySQL sample for replay regression
Branch: feat/nethttp-mysql-hotloop-regression

Adds a minimal net/http + database/sql + go-sql-driver/mysql sample that exercises the full MySQL connection and command phase end-to-end. Shape of the app: - /health issues db.Ping() (forces a real command roundtrip). - /users SELECTs from a seeded table. - /users/add INSERTs and returns the row. Why: Keploy's OSS pipeline consumes this sample as a regression guard for the MySQL replay command-phase read deadline. If the proxy ever re-introduces a zero or negative read deadline, the Go driver blocks inside db.Ping() during replay and the HTTP listener never binds :8080, making the regression loud in CI rather than silently hanging. Signed-off-by: Asish Kumar <officialasishkumar@gmail.com>

Copilot

Pull request overview

Adds a new nethttp-mysql sample app (net/http + database/sql + go-sql-driver/mysql) intended to act as a CI regression guard for MySQL replay command-phase read-deadline issues by ensuring db.Ping() completes and the server successfully binds :8080.

Changes:

Introduces a minimal HTTP server with /health, /users, and /users/add backed by a MySQL users table.
Adds module files (go.mod/go.sum) for the new sample.
Adds Dockerfile + sample README with local-run and Keploy record/test instructions.

Reviewed changes

Copilot reviewed 4 out of 5 changed files in this pull request and generated 8 comments.

Show a summary per file

File	Description
nethttp-mysql/main.go	New MySQL-backed HTTP server + startup schema/seed logic used to exercise command-phase traffic.
nethttp-mysql/README.md	Local Docker + Keploy record/test instructions for the regression sample.
nethttp-mysql/Dockerfile	Container build/runtime packaging for the sample.
nethttp-mysql/go.mod	Declares the sample’s standalone Go module and MySQL driver dependency.
nethttp-mysql/go.sum	Locks dependency checksums for reproducible builds.

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

Copilot · 2026-04-21T15:25:52Z

+	for attempt := 1; attempt <= 30; attempt++ {
+		db, err = sql.Open("mysql", dsn)
+		if err == nil {
+			err = db.Ping()
+			if err == nil {
+				break
+			}
+		}
+		log.Printf("waiting for mysql (attempt %d): %v", attempt, err)
+		time.Sleep(2 * time.Second)
+	}
+	if err != nil {
+		log.Fatalf("could not connect to mysql after retries: %v", err)


In the retry loop, a new *sql.DB handle is created on every attempt, but failed attempts aren't closed. If Ping() fails, the previous handle can keep underlying connections/resources around. Consider calling Close() before reassigning, or Open() once and just retry Ping() on the same handle.

Suggested change

for attempt := 1; attempt <= 30; attempt++ {

db, err = sql.Open("mysql", dsn)

if err == nil {

err = db.Ping()

if err == nil {

break

}

}

log.Printf("waiting for mysql (attempt %d): %v", attempt, err)

time.Sleep(2 * time.Second)

}

if err != nil {

log.Fatalf("could not connect to mysql after retries: %v", err)

db, err = sql.Open("mysql", dsn)

if err != nil {

log.Fatalf("could not create mysql handle: %v; verify the DSN and MySQL driver configuration, then restart the service", err)

}

defer db.Close()

for attempt := 1; attempt <= 30; attempt++ {

err = db.Ping()

if err == nil {

break

}

log.Printf("waiting for mysql (attempt %d): %v; ensure the MySQL container is running and reachable at %s:%s, then retry", attempt, err, host, port)

time.Sleep(2 * time.Second)

}

if err != nil {

log.Fatalf("could not connect to mysql after retries: %v; confirm MySQL is healthy and the connection settings are correct, then restart the service", err)

Copilot · 2026-04-21T15:25:53Z

+	}
+
+	var count int
+	_ = db.QueryRow("SELECT COUNT(*) FROM users").Scan(&count)


The result of QueryRow(...).Scan(&count) is intentionally ignored. If this query fails (e.g., missing permissions/table issues), count stays 0 and the app will try to seed, potentially masking the root cause. Please handle the Scan error and fail fast (or at least log it) so startup is deterministic.

Suggested change

_ = db.QueryRow("SELECT COUNT(*) FROM users").Scan(&count)

if err := db.QueryRow("SELECT COUNT(*) FROM users").Scan(&count); err != nil {

log.Fatalf("failed to count existing users; verify the users table exists and the configured MySQL user has SELECT access: %v", err)

}

Copilot · 2026-04-21T15:25:53Z

+	}
+	w.Header().Set("Content-Type", "application/json")
+	_ = json.NewEncoder(w).Encode(users)
+}


After iterating rows.Next(), rows.Err() is not checked, and JSON encoding errors are ignored. This can cause partial/empty responses to be returned as 200 OK even when iteration/encoding fails. Please check rows.Err() after the loop and handle/return any Encode error.

Copilot · 2026-04-21T15:25:53Z

+		http.Error(w, err.Error(), http.StatusInternalServerError)
+		return
+	}
+	id, _ := res.LastInsertId()


res.LastInsertId() error is ignored. If the driver doesn't support it or it fails, the handler will silently return id=0. Please handle the error and return a 500 (or similar) if the insert ID can't be retrieved.

Suggested change

id, _ := res.LastInsertId()

id, err := res.LastInsertId()

if err != nil {

http.Error(w, "failed to retrieve the inserted user ID; please retry or verify the database driver supports LastInsertId", http.StatusInternalServerError)

return

}

Copilot · 2026-04-21T15:25:54Z

+		time.Sleep(2 * time.Second)
+	}
+	if err != nil {
+		log.Fatalf("could not connect to mysql after retries: %v", err)


This fatal error message doesn't give a clear next step for debugging (e.g., which env vars/host were used, or suggesting to check the MySQL container/network). Consider including the target host:port/db and a suggested action (verify MySQL is running/reachable and credentials are correct) to make CI/local failures faster to diagnose.

Suggested change

log.Fatalf("could not connect to mysql after retries: %v", err)

log.Fatalf("could not connect to MySQL after retries (target=%s:%s/%s): %v; verify MySQL is running and reachable at that address and that MYSQL_USER/MYSQL_PASSWORD/MYSQL_DATABASE are correct", host, port, dbname, err)

Copilot · 2026-04-21T15:25:54Z

+RUN go mod download || true
+COPY . .
+RUN go mod tidy && CGO_ENABLED=0 GOOS=linux go build -o app .


go mod download || true will hide real module download failures and can make builds succeed with missing deps until a later step fails. Also, running go mod tidy during image build makes the build non-deterministic and can mutate the module graph at build time. Prefer failing fast on go mod download and running go build without go mod tidy here (keep tidy as a repo-side step).

Suggested change

RUN go mod download || true

COPY . .

RUN go mod tidy && CGO_ENABLED=0 GOOS=linux go build -o app .

RUN go mod download

COPY . .

RUN CGO_ENABLED=0 GOOS=linux go build -o app .

Copilot · 2026-04-21T15:25:54Z

+keploy test -c "docker run -p 8080:8080 --name nethttp-mysql --network keploy-network \
+  -e MYSQL_HOST=mysql -e MYSQL_USER=root -e MYSQL_PASSWORD=password \
+  -e MYSQL_DATABASE=testdb nethttp-mysql" \
+  --containerName nethttp-mysql --network-name keploy-network \


The Keploy CLI flag here uses --containerName, but other samples in this repo consistently use --container-name (e.g., http-postgres/README.md:58, fasthttp-postgres/README.md:25). If --containerName isn't a valid alias, the command will fail. Please align this to --container-name for consistency and correctness.

Suggested change

--containerName nethttp-mysql --network-name keploy-network \

--container-name nethttp-mysql --network-name keploy-network \

Copilot · 2026-04-21T15:25:55Z

+}
+
+func getUsers(w http.ResponseWriter, r *http.Request) {
+	rows, err := db.Query("SELECT id, name, email FROM users")


The /users query has no ORDER BY, so row order is not guaranteed and can change as the table grows/changes. For a sample used in record/replay regression, this can introduce flaky diffs. Consider adding an explicit ordering (e.g., by id) to keep responses deterministic.

Suggested change

rows, err := db.Query("SELECT id, name, email FROM users")

rows, err := db.Query("SELECT id, name, email FROM users ORDER BY id")

The e2e pipeline cannot discriminate a fix from a regression for this specific bug. Triggering the zero/overflow read-deadline path requires SQLDelay = 0 (or a real seconds-valued Duration) in the replay command-phase loop. SQLDelay is derived exclusively from Test.Delay via `time.Duration(Test.Delay) * time.Second`, and the same value gates keploy's app-startup wait before replay begins. Pushing Test.Delay = 0 into the driver script to exercise the zero-deadline path also causes keploy to replay before the sample binds :8080, so the matrix fails for the wrong reason on both buggy and fixed binaries. With any non-zero Test.Delay the two pre-fix unit bugs (`time.Duration(Test.Delay)` without `* time.Second` in the caller and `2 * time.Second * time.Duration(opts.SQLDelay)` in the proxy) cancel out and the buggy released binary computes a valid positive deadline. Net: all three matrix cells would stay green regardless of whether the fix is present, so the entry provides no signal. Drop the matrix entry, the driver script, and the prepare_and_run trigger change that was only there to let this stacked branch run outside main. The fix remains guarded by the derivation unit test in query_readtimeout_test.go, which pins the math across the zero, sub-second, seconds, and hours cases. The nethttp-mysql sample in keploy/samples-go#224 is unaffected and stays in place as a general MySQL coverage sample. - Remove nethttp-mysql entry from .github/workflows/golang_docker.yml. - Delete .github/workflows/test_workflow_scripts/golang/nethttp_mysql/ golang-docker.sh. - Restore the main-only branch filter on pull_request and push in .github/workflows/prepare_and_run.yml. Signed-off-by: Asish Kumar <officialasishkumar@gmail.com>

…se (#4095) * fix(mysql): prevent zero/overflow read deadline in replay command phase The command-phase loop in pkg/agent/proxy/integrations/mysql/replayer/ query.go derived its per-iteration read deadline as readTimeout := 2 * time.Second * time.Duration(opts.SQLDelay) which was wrong in two directions: - When the HTTP caller sent SQLDelay=0 (a legitimate shape once the field is serialized across the agent/client boundary), the product was 0s. SetReadDeadline(time.Now().Add(0)) set an already-expired deadline, ReadPacketBuffer returned a timeout without reading any bytes, and the timeout branch's 50ms backoff became a hot loop that never consumed the client's MySQL commands. The client driver sat in a blocking Read waiting for a response that would never come. - When SQLDelay was a genuine seconds-valued time.Duration (e.g. 5s = 5e9), the product 2 * 1e9 * 5e9 = 1e19 overflowed int64, wrapped negative, and produced the same expired-deadline hot loop from the opposite end. The derivation is replaced with readTimeout := 2 * opts.SQLDelay if readTimeout < time.Second { readTimeout = 2 * time.Second } which treats SQLDelay as a duration (matching how the rest of the codebase uses time.Duration) and guarantees a safe positive deadline even when the caller sends zero. The two callers that build OutgoingOptions (pkg/service/replay and pkg/service/runner) were also feeding r.config.Test.Delay (a uint64 meaning seconds) through time.Duration() without multiplying by time.Second, producing a nanosecond-valued Duration. Those are corrected to `time.Duration(Delay) * time.Second` so the knob's unit matches user intent and the startup-wait path (which already uses `* time.Second`). Regression guard: a new golang_docker matrix entry (nethttp-mysql) drives a net/http + database/sql sample end-to-end. If this bug ever re-lands, the sample's db.Ping() will block during replay, the HTTP server will not bind :8080, and every replayed test case will fail with connection reset — making the regression loud in CI. A small unit test in query_readtimeout_test.go pins the derivation math itself so future refactors cannot silently reintroduce a zero or sub-second deadline. Signed-off-by: Asish Kumar <officialasishkumar@gmail.com> * ci: remove main-only branch filter from prepare_and_run triggers - Drops `branches: [main]` from both the pull_request and push triggers in prepare_and_run.yml so the added dns_strict_resolver and nethttp-mysql pipelines (plus the rest of the umbrella) run on PRs and pushes targeting any base branch, not just main. Signed-off-by: Asish Kumar <officialasishkumar@gmail.com> * chore: retrigger CI pipeline - Empty commit to retrigger the CI pipeline. Signed-off-by: Asish Kumar <officialasishkumar@gmail.com> * ci: drop nethttp-mysql e2e wiring for this fix The e2e pipeline cannot discriminate a fix from a regression for this specific bug. Triggering the zero/overflow read-deadline path requires SQLDelay = 0 (or a real seconds-valued Duration) in the replay command-phase loop. SQLDelay is derived exclusively from Test.Delay via `time.Duration(Test.Delay) * time.Second`, and the same value gates keploy's app-startup wait before replay begins. Pushing Test.Delay = 0 into the driver script to exercise the zero-deadline path also causes keploy to replay before the sample binds :8080, so the matrix fails for the wrong reason on both buggy and fixed binaries. With any non-zero Test.Delay the two pre-fix unit bugs (`time.Duration(Test.Delay)` without `* time.Second` in the caller and `2 * time.Second * time.Duration(opts.SQLDelay)` in the proxy) cancel out and the buggy released binary computes a valid positive deadline. Net: all three matrix cells would stay green regardless of whether the fix is present, so the entry provides no signal. Drop the matrix entry, the driver script, and the prepare_and_run trigger change that was only there to let this stacked branch run outside main. The fix remains guarded by the derivation unit test in query_readtimeout_test.go, which pins the math across the zero, sub-second, seconds, and hours cases. The nethttp-mysql sample in keploy/samples-go#224 is unaffected and stays in place as a general MySQL coverage sample. - Remove nethttp-mysql entry from .github/workflows/golang_docker.yml. - Delete .github/workflows/test_workflow_scripts/golang/nethttp_mysql/ golang-docker.sh. - Restore the main-only branch filter on pull_request and push in .github/workflows/prepare_and_run.yml. Signed-off-by: Asish Kumar <officialasishkumar@gmail.com> * chore: retrigger CI pipeline - Empty commit to kick off the workflow after rebasing the branch onto main. Signed-off-by: Asish Kumar <officialasishkumar@gmail.com> --------- Signed-off-by: Asish Kumar <officialasishkumar@gmail.com>

Copilot AI review requested due to automatic review settings April 21, 2026 15:20

officialasishkumar requested a review from khareyash05 as a code owner April 21, 2026 15:20

Copilot started reviewing on behalf of officialasishkumar April 21, 2026 15:21 View session

officialasishkumar mentioned this pull request Apr 21, 2026

fix(mysql): prevent zero/overflow read deadline in replay command phase keploy/keploy#4095

Merged

2 tasks

Copilot AI reviewed Apr 21, 2026

View reviewed changes

officialasishkumar mentioned this pull request Apr 22, 2026

[bug]: MySQL replay command-phase hot-loop on zero/overflow read deadline keploy/keploy#4102

Closed

1 task

khareyash05 approved these changes Apr 23, 2026

View reviewed changes

khareyash05 merged commit 3ef38be into main Apr 23, 2026
77 checks passed

khareyash05 deleted the feat/nethttp-mysql-hotloop-regression branch April 23, 2026 12:04

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

feat(nethttp-mysql): add net/http + MySQL sample for replay regression#224

feat(nethttp-mysql): add net/http + MySQL sample for replay regression#224
khareyash05 merged 1 commit intomainfrom
feat/nethttp-mysql-hotloop-regression

officialasishkumar commented Apr 21, 2026 •

edited

Loading

Uh oh!

Copilot AI left a comment

Uh oh!

Copilot AI Apr 21, 2026

Uh oh!

Copilot AI Apr 21, 2026

Uh oh!

Copilot AI Apr 21, 2026

Uh oh!

Copilot AI Apr 21, 2026

Uh oh!

Copilot AI Apr 21, 2026

Uh oh!

Copilot AI Apr 21, 2026

Uh oh!

Copilot AI Apr 21, 2026

Uh oh!

Copilot AI Apr 21, 2026

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

-	_ = db.QueryRow("SELECT COUNT(*) FROM users").Scan(&count)
+	if err := db.QueryRow("SELECT COUNT(*) FROM users").Scan(&count); err != nil {
+		log.Fatalf("failed to count existing users; verify the users table exists and the configured MySQL user has SELECT access: %v", err)
+	}

	log.Fatalf("could not connect to mysql after retries: %v", err)
	log.Fatalf("could not connect to MySQL after retries (target=%s:%s/%s): %v; verify MySQL is running and reachable at that address and that MYSQL_USER/MYSQL_PASSWORD/MYSQL_DATABASE are correct", host, port, dbname, err)

	--containerName nethttp-mysql --network-name keploy-network \
	--container-name nethttp-mysql --network-name keploy-network \

	rows, err := db.Query("SELECT id, name, email FROM users")
	rows, err := db.Query("SELECT id, name, email FROM users ORDER BY id")

Conversation

officialasishkumar commented Apr 21, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Describe the changes that are made

Links & References

🔗 Related PRs

🐞 Related Issues

📄 Related Documents

What type of PR is this? (check all applicable)

Added e2e test pipeline?

Added comments for hard-to-understand areas?

Added to documentation?

Are there any sample code or steps to test the changes?

Self Review done?

Any relevant screenshots, recordings or logs?

🧠 PR Semantics

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Reviewed changes

Uh oh!

Copilot AI Apr 21, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Apr 21, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Apr 21, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Apr 21, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Apr 21, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Apr 21, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Apr 21, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Apr 21, 2026

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

officialasishkumar commented Apr 21, 2026 •

edited

Loading