shishantbiswas/coder
1
0
Fork 0
mirror of https://github.com/coder/coder.git synced 2026-06-02 20:48:20 +00:00
Code Issues Packages Projects Releases Wiki Activity

chore: improve healthcheck timeout message (#21520)

Browse Source 
Relates to https://github.com/coder/internal/issues/272

This flake has been persisting for a while, and unfortunately there's no
detail on which healthcheck in particular is holding things up.

This PR adds a concurrency-safe `healthcheck.Progress` and wires it
through `healthcheck.Run`. If the healthcheck times out, it will provide
information on which healthchecks are completed / running, and how long
they took / are still taking.

🤖 Claude Opus 4.5 completed the first round of this implementation,
which I then refactored.
This commit is contained in:
Cian Johnston
2026-01-15 16:37:05 +00:00
committed by GitHub
parent 7fc84ecf0b
commit 3a62a8e70e
6 changed files with 216 additions and 24 deletions
Download Patch File Download Diff File Expand all files Collapse all files
coderd/debug_test.go
+20 -17
View File
@@ -14,6 +14,8 @@ import (
"cdr.dev/slog/v3/sloggers/slogtest"
"github.com/coder/coder/v2/coderd/coderdtest"
"github.com/coder/coder/v2/coderd/healthcheck"
"github.com/coder/coder/v2/codersdk"
"github.com/coder/coder/v2/codersdk/healthsdk"
"github.com/coder/coder/v2/testutil"
)
@@ -28,7 +30,7 @@ func TestDebugHealth(t *testing.T) {
ctx, cancel = context.WithTimeout(context.Background(), testutil.WaitShort)
sessionToken string
client = coderdtest.New(t, &coderdtest.Options{
HealthcheckFunc: func(_ context.Context, apiKey string) *healthsdk.HealthcheckReport {
HealthcheckFunc: func(_ context.Context, apiKey string, _ *healthcheck.Progress) *healthsdk.HealthcheckReport {
calls.Add(1)
assert.Equal(t, sessionToken, apiKey)
return &healthsdk.HealthcheckReport{
@@ -61,7 +63,7 @@ func TestDebugHealth(t *testing.T) {
ctx, cancel = context.WithTimeout(context.Background(), testutil.WaitShort)
sessionToken string
client = coderdtest.New(t, &coderdtest.Options{
HealthcheckFunc: func(_ context.Context, apiKey string) *healthsdk.HealthcheckReport {
HealthcheckFunc: func(_ context.Context, apiKey string, _ *healthcheck.Progress) *healthsdk.HealthcheckReport {
calls.Add(1)
assert.Equal(t, sessionToken, apiKey)
return &healthsdk.HealthcheckReport{
@@ -93,19 +95,14 @@ func TestDebugHealth(t *testing.T) {
// Need to ignore errors due to ctx timeout
logger = slogtest.Make(t, &slogtest.Options{IgnoreErrors: true})
ctx, cancel = context.WithTimeout(context.Background(), testutil.WaitShort)
done = make(chan struct{})
client = coderdtest.New(t, &coderdtest.Options{
Logger: &logger,
HealthcheckTimeout: time.Microsecond,
HealthcheckFunc: func(context.Context, string) *healthsdk.HealthcheckReport {
t := time.NewTimer(time.Second)
defer t.Stop()
select {
case <-ctx.Done():
return &healthsdk.HealthcheckReport{}
case <-t.C:
return &healthsdk.HealthcheckReport{}
}
HealthcheckTimeout: time.Second,
HealthcheckFunc: func(_ context.Context, _ string, progress *healthcheck.Progress) *healthsdk.HealthcheckReport {
progress.Start("test")
<-done
return &healthsdk.HealthcheckReport{}
},
})
_ = coderdtest.CreateFirstUser(t, client)
@@ -115,8 +112,14 @@ func TestDebugHealth(t *testing.T) {
res, err := client.Request(ctx, "GET", "/api/v2/debug/health", nil)
require.NoError(t, err)
defer res.Body.Close()
_, _ = io.ReadAll(res.Body)
close(done)
bs, err := io.ReadAll(res.Body)
require.NoError(t, err, "reading body")
require.Equal(t, http.StatusServiceUnavailable, res.StatusCode)
var sdkResp codersdk.Response
require.NoError(t, json.Unmarshal(bs, &sdkResp), "unmarshaling sdk response")
require.Equal(t, "Healthcheck timed out.", sdkResp.Message)
require.Contains(t, sdkResp.Detail, "Still running: test (elapsed:")
})
t.Run("Refresh", func(t *testing.T) {
@@ -128,7 +131,7 @@ func TestDebugHealth(t *testing.T) {
ctx, cancel = context.WithTimeout(context.Background(), testutil.WaitShort)
client = coderdtest.New(t, &coderdtest.Options{
HealthcheckRefresh: time.Microsecond,
HealthcheckFunc: func(context.Context, string) *healthsdk.HealthcheckReport {
HealthcheckFunc: func(context.Context, string, *healthcheck.Progress) *healthsdk.HealthcheckReport {
calls <- struct{}{}
return &healthsdk.HealthcheckReport{}
},
@@ -173,7 +176,7 @@ func TestDebugHealth(t *testing.T) {
client = coderdtest.New(t, &coderdtest.Options{
HealthcheckRefresh: time.Hour,
HealthcheckTimeout: time.Hour,
HealthcheckFunc: func(context.Context, string) *healthsdk.HealthcheckReport {
HealthcheckFunc: func(context.Context, string, *healthcheck.Progress) *healthsdk.HealthcheckReport {
calls++
return &healthsdk.HealthcheckReport{
Time: time.Now(),
@@ -207,7 +210,7 @@ func TestDebugHealth(t *testing.T) {
ctx, cancel = context.WithTimeout(context.Background(), testutil.WaitShort)
sessionToken string
client = coderdtest.New(t, &coderdtest.Options{
HealthcheckFunc: func(_ context.Context, apiKey string) *healthsdk.HealthcheckReport {
HealthcheckFunc: func(_ context.Context, apiKey string, _ *healthcheck.Progress) *healthsdk.HealthcheckReport {
assert.Equal(t, sessionToken, apiKey)
return &healthsdk.HealthcheckReport{
Time: time.Now(),