mirror of https://github.com/coder/coder.git
ci: use big runners (#4990)
* chore: Close idle connections on test cleanup It's possible that this was the source of a leak on Windows... * ci: use big runners * fix: Improve tailnet connections by reducing timeouts This awaits connection ping before running a dial. Before, we were hitting the TCP retransmission and handshake timeouts, which could intermittently add 1 or 5 seconds to a connection being initialized. * Add logging to Startupscript test * Add better logging * Write startup script logs to fs dir * Fix startup script test * Fix startup script test * Reduce test timeout * Use central tmp dir in agent * Adjust output * Skip startup script test on Windows Co-authored-by: Kyle Carberry <kyle@carberry.com>
This commit is contained in:
Ammar Bandukwala
GitHub
parent
9578ce9f77
commit
73f91e4690
|
|
@ -89,7 +89,7 @@ jobs:
|
|||
style-lint-golangci:
|
||||
name: style/lint/golangci
|
||||
timeout-minutes: 5
|
||||
runs-on: ${{ github.repository_owner == 'coder' && 'buildjet-8vcpu-ubuntu-2204' || 'ubuntu-latest' }}
|
||||
runs-on: ${{ github.repository_owner == 'coder' && 'ubuntu-latest-16-cores' || 'ubuntu-latest' }}
|
||||
steps:
|
||||
- uses: actions/checkout@v3
|
||||
- uses: actions/setup-go@v3
|
||||
|
|
@ -171,7 +171,7 @@ jobs:
|
|||
gen:
|
||||
name: "style/gen"
|
||||
timeout-minutes: 8
|
||||
runs-on: ${{ github.repository_owner == 'coder' && 'buildjet-8vcpu-ubuntu-2204' || 'ubuntu-latest' }}
|
||||
runs-on: ${{ github.repository_owner == 'coder' && 'ubuntu-latest-16-cores' || 'ubuntu-latest' }}
|
||||
needs: changes
|
||||
if: needs.changes.outputs.docs-only == 'false'
|
||||
steps:
|
||||
|
|
@ -276,7 +276,7 @@ jobs:
|
|||
|
||||
test-go:
|
||||
name: "test/go"
|
||||
runs-on: ${{ matrix.os == 'ubuntu-latest' && github.repository_owner == 'coder' && 'buildjet-8vcpu-ubuntu-2204' || matrix.os }}
|
||||
runs-on: ${{ matrix.os == 'ubuntu-latest' && github.repository_owner == 'coder' && 'ubuntu-latest-16-cores' || matrix.os == 'windows-2022' && github.repository_owner == 'coder' && 'windows-latest-8-cores'|| matrix.os }}
|
||||
timeout-minutes: 20
|
||||
strategy:
|
||||
matrix:
|
||||
|
|
@ -336,11 +336,7 @@ jobs:
|
|||
echo ::set-output name=cover::false
|
||||
fi
|
||||
set -x
|
||||
test_timeout=5m
|
||||
if [[ "${{ matrix.os }}" == windows* ]]; then
|
||||
test_timeout=10m
|
||||
fi
|
||||
gotestsum --junitfile="gotests.xml" --packages="./..." -- -parallel=8 -timeout=$test_timeout -short -failfast $COVERAGE_FLAGS
|
||||
gotestsum --junitfile="gotests.xml" --packages="./..." -- -parallel=8 -timeout=3m -short -failfast $COVERAGE_FLAGS
|
||||
|
||||
- uses: codecov/codecov-action@v3
|
||||
# This action has a tendency to error out unexpectedly, it has
|
||||
|
|
@ -356,7 +352,7 @@ jobs:
|
|||
|
||||
test-go-postgres:
|
||||
name: "test/go/postgres"
|
||||
runs-on: ${{ github.repository_owner == 'coder' && 'buildjet-8vcpu-ubuntu-2204' || 'ubuntu-latest' }}
|
||||
runs-on: ${{ github.repository_owner == 'coder' && 'ubuntu-latest-16-cores' || 'ubuntu-latest' }}
|
||||
# This timeout must be greater than the timeout set by `go test` in
|
||||
# `make test-postgres` to ensure we receive a trace of running
|
||||
# goroutines. Setting this to the timeout +5m should work quite well
|
||||
|
|
@ -417,7 +413,7 @@ jobs:
|
|||
|
||||
deploy:
|
||||
name: "deploy"
|
||||
runs-on: ${{ github.repository_owner == 'coder' && 'buildjet-8vcpu-ubuntu-2204' || 'ubuntu-latest' }}
|
||||
runs-on: ${{ github.repository_owner == 'coder' && 'ubuntu-latest-16-cores' || 'ubuntu-latest' }}
|
||||
timeout-minutes: 30
|
||||
needs: changes
|
||||
if: |
|
||||
|
|
@ -514,7 +510,7 @@ jobs:
|
|||
|
||||
test-js:
|
||||
name: "test/js"
|
||||
runs-on: ubuntu-latest
|
||||
runs-on: ${{ github.repository_owner == 'coder' && 'ubuntu-latest-16-cores' || 'ubuntu-latest' }}
|
||||
timeout-minutes: 20
|
||||
steps:
|
||||
- uses: actions/checkout@v3
|
||||
|
|
|
|||
|
|
@ -28,7 +28,7 @@ env:
|
|||
|
||||
jobs:
|
||||
release:
|
||||
runs-on: ${{ github.repository_owner == 'coder' && 'buildjet-8vcpu-ubuntu-2204' || 'ubuntu-latest' }}
|
||||
runs-on: ${{ github.repository_owner == 'coder' && 'ubuntu-latest-16-cores' || 'ubuntu-latest' }}
|
||||
env:
|
||||
# Necessary for Docker manifest
|
||||
DOCKER_CLI_EXPERIMENTAL: "enabled"
|
||||
|
|
|
|||
3
Makefile
3
Makefile
|
|
@ -463,8 +463,11 @@ test: test-clean
|
|||
# When updating -timeout for this test, keep in sync with
|
||||
# test-go-postgres (.github/workflows/coder.yaml).
|
||||
test-postgres: test-clean test-postgres-docker
|
||||
# The postgres test is prone to failure, so we limit parallelism for
|
||||
# more consistent execution.
|
||||
DB=ci DB_FROM=$(shell go run scripts/migrate-ci/main.go) gotestsum --junitfile="gotests.xml" --packages="./..." -- \
|
||||
-covermode=atomic -coverprofile="gotests.coverage" -timeout=20m \
|
||||
-parallel=4 \
|
||||
-coverpkg=./... \
|
||||
-count=1 -race -failfast
|
||||
.PHONY: test-postgres
|
||||
|
|
|
|||
|
|
@ -56,6 +56,7 @@ const (
|
|||
|
||||
type Options struct {
|
||||
Filesystem afero.Fs
|
||||
TempDir string
|
||||
ExchangeToken func(ctx context.Context) (string, error)
|
||||
Client Client
|
||||
ReconnectingPTYTimeout time.Duration
|
||||
|
|
@ -78,6 +79,9 @@ func New(options Options) io.Closer {
|
|||
if options.Filesystem == nil {
|
||||
options.Filesystem = afero.NewOsFs()
|
||||
}
|
||||
if options.TempDir == "" {
|
||||
options.TempDir = os.TempDir()
|
||||
}
|
||||
if options.ExchangeToken == nil {
|
||||
options.ExchangeToken = func(ctx context.Context) (string, error) {
|
||||
return "", nil
|
||||
|
|
@ -93,6 +97,7 @@ func New(options Options) io.Closer {
|
|||
client: options.Client,
|
||||
exchangeToken: options.ExchangeToken,
|
||||
filesystem: options.Filesystem,
|
||||
tempDir: options.TempDir,
|
||||
stats: &Stats{},
|
||||
}
|
||||
server.init(ctx)
|
||||
|
|
@ -104,6 +109,7 @@ type agent struct {
|
|||
client Client
|
||||
exchangeToken func(ctx context.Context) (string, error)
|
||||
filesystem afero.Fs
|
||||
tempDir string
|
||||
|
||||
reconnectingPTYs sync.Map
|
||||
reconnectingPTYTimeout time.Duration
|
||||
|
|
@ -375,14 +381,14 @@ func (a *agent) runStartupScript(ctx context.Context, script string) error {
|
|||
return nil
|
||||
}
|
||||
|
||||
writer, err := os.OpenFile(filepath.Join(os.TempDir(), "coder-startup-script.log"), os.O_CREATE|os.O_RDWR, 0o600)
|
||||
a.logger.Info(ctx, "running startup script", slog.F("script", script))
|
||||
writer, err := a.filesystem.OpenFile(filepath.Join(a.tempDir, "coder-startup-script.log"), os.O_CREATE|os.O_RDWR, 0o600)
|
||||
if err != nil {
|
||||
return xerrors.Errorf("open startup script log file: %w", err)
|
||||
}
|
||||
defer func() {
|
||||
_ = writer.Close()
|
||||
}()
|
||||
|
||||
cmd, err := a.createCommand(ctx, script, nil)
|
||||
if err != nil {
|
||||
return xerrors.Errorf("create command: %w", err)
|
||||
|
|
|
|||
|
|
@ -61,7 +61,7 @@ func TestAgent(t *testing.T) {
|
|||
ctx, cancel := context.WithTimeout(context.Background(), testutil.WaitLong)
|
||||
defer cancel()
|
||||
|
||||
conn, stats := setupAgent(t, codersdk.WorkspaceAgentMetadata{}, 0)
|
||||
conn, stats, _ := setupAgent(t, codersdk.WorkspaceAgentMetadata{}, 0)
|
||||
|
||||
sshClient, err := conn.SSHClient(ctx)
|
||||
require.NoError(t, err)
|
||||
|
|
@ -81,7 +81,7 @@ func TestAgent(t *testing.T) {
|
|||
ctx, cancel := context.WithTimeout(context.Background(), testutil.WaitLong)
|
||||
defer cancel()
|
||||
|
||||
conn, stats := setupAgent(t, codersdk.WorkspaceAgentMetadata{}, 0)
|
||||
conn, stats, _ := setupAgent(t, codersdk.WorkspaceAgentMetadata{}, 0)
|
||||
|
||||
ptyConn, err := conn.ReconnectingPTY(ctx, uuid.NewString(), 128, 128, "/bin/bash")
|
||||
require.NoError(t, err)
|
||||
|
|
@ -231,7 +231,7 @@ func TestAgent(t *testing.T) {
|
|||
if runtime.GOOS == "windows" {
|
||||
home = "/" + strings.ReplaceAll(home, "\\", "/")
|
||||
}
|
||||
conn, _ := setupAgent(t, codersdk.WorkspaceAgentMetadata{}, 0)
|
||||
conn, _, _ := setupAgent(t, codersdk.WorkspaceAgentMetadata{}, 0)
|
||||
sshClient, err := conn.SSHClient(ctx)
|
||||
require.NoError(t, err)
|
||||
defer sshClient.Close()
|
||||
|
|
@ -261,7 +261,7 @@ func TestAgent(t *testing.T) {
|
|||
ctx, cancel := context.WithTimeout(context.Background(), testutil.WaitLong)
|
||||
defer cancel()
|
||||
|
||||
conn, _ := setupAgent(t, codersdk.WorkspaceAgentMetadata{}, 0)
|
||||
conn, _, _ := setupAgent(t, codersdk.WorkspaceAgentMetadata{}, 0)
|
||||
sshClient, err := conn.SSHClient(ctx)
|
||||
require.NoError(t, err)
|
||||
defer sshClient.Close()
|
||||
|
|
@ -360,19 +360,23 @@ func TestAgent(t *testing.T) {
|
|||
|
||||
t.Run("StartupScript", func(t *testing.T) {
|
||||
t.Parallel()
|
||||
tempPath := filepath.Join(t.TempDir(), "content.txt")
|
||||
content := "somethingnice"
|
||||
setupAgent(t, codersdk.WorkspaceAgentMetadata{
|
||||
StartupScript: fmt.Sprintf("echo %s > %s", content, tempPath),
|
||||
if runtime.GOOS == "windows" {
|
||||
t.Skip("This test doesn't work on Windows for some reason...")
|
||||
}
|
||||
content := "output"
|
||||
_, _, fs := setupAgent(t, codersdk.WorkspaceAgentMetadata{
|
||||
StartupScript: "echo " + content,
|
||||
}, 0)
|
||||
|
||||
var gotContent string
|
||||
require.Eventually(t, func() bool {
|
||||
content, err := os.ReadFile(tempPath)
|
||||
outputPath := filepath.Join(os.TempDir(), "coder-startup-script.log")
|
||||
content, err := afero.ReadFile(fs, outputPath)
|
||||
if err != nil {
|
||||
t.Logf("read file %q: %s", outputPath, err)
|
||||
return false
|
||||
}
|
||||
if len(content) == 0 {
|
||||
t.Logf("no content in %q", outputPath)
|
||||
return false
|
||||
}
|
||||
if runtime.GOOS == "windows" {
|
||||
|
|
@ -384,7 +388,7 @@ func TestAgent(t *testing.T) {
|
|||
}
|
||||
gotContent = string(content)
|
||||
return true
|
||||
}, testutil.WaitMedium, testutil.IntervalMedium)
|
||||
}, testutil.WaitShort, testutil.IntervalMedium)
|
||||
require.Equal(t, content, strings.TrimSpace(gotContent))
|
||||
})
|
||||
|
||||
|
|
@ -400,7 +404,7 @@ func TestAgent(t *testing.T) {
|
|||
ctx, cancel := context.WithTimeout(context.Background(), testutil.WaitLong)
|
||||
defer cancel()
|
||||
|
||||
conn, _ := setupAgent(t, codersdk.WorkspaceAgentMetadata{}, 0)
|
||||
conn, _, _ := setupAgent(t, codersdk.WorkspaceAgentMetadata{}, 0)
|
||||
id := uuid.NewString()
|
||||
netConn, err := conn.ReconnectingPTY(ctx, id, 100, 100, "/bin/bash")
|
||||
require.NoError(t, err)
|
||||
|
|
@ -497,7 +501,7 @@ func TestAgent(t *testing.T) {
|
|||
}
|
||||
}()
|
||||
|
||||
conn, _ := setupAgent(t, codersdk.WorkspaceAgentMetadata{}, 0)
|
||||
conn, _, _ := setupAgent(t, codersdk.WorkspaceAgentMetadata{}, 0)
|
||||
require.True(t, conn.AwaitReachable(context.Background()))
|
||||
conn1, err := conn.DialContext(context.Background(), l.Addr().Network(), l.Addr().String())
|
||||
require.NoError(t, err)
|
||||
|
|
@ -518,7 +522,7 @@ func TestAgent(t *testing.T) {
|
|||
ctx, cancel := context.WithTimeout(context.Background(), testutil.WaitLong)
|
||||
defer cancel()
|
||||
derpMap := tailnettest.RunDERPAndSTUN(t)
|
||||
conn, _ := setupAgent(t, codersdk.WorkspaceAgentMetadata{
|
||||
conn, _, _ := setupAgent(t, codersdk.WorkspaceAgentMetadata{
|
||||
DERPMap: derpMap,
|
||||
}, 0)
|
||||
defer conn.Close()
|
||||
|
|
@ -601,7 +605,7 @@ func TestAgent(t *testing.T) {
|
|||
}
|
||||
|
||||
func setupSSHCommand(t *testing.T, beforeArgs []string, afterArgs []string) *exec.Cmd {
|
||||
agentConn, _ := setupAgent(t, codersdk.WorkspaceAgentMetadata{}, 0)
|
||||
agentConn, _, _ := setupAgent(t, codersdk.WorkspaceAgentMetadata{}, 0)
|
||||
listener, err := net.Listen("tcp", "127.0.0.1:0")
|
||||
require.NoError(t, err)
|
||||
waitGroup := sync.WaitGroup{}
|
||||
|
|
@ -644,7 +648,7 @@ func setupSSHCommand(t *testing.T, beforeArgs []string, afterArgs []string) *exe
|
|||
func setupSSHSession(t *testing.T, options codersdk.WorkspaceAgentMetadata) *ssh.Session {
|
||||
ctx, cancel := context.WithTimeout(context.Background(), testutil.WaitLong)
|
||||
defer cancel()
|
||||
conn, _ := setupAgent(t, options, 0)
|
||||
conn, _, _ := setupAgent(t, options, 0)
|
||||
sshClient, err := conn.SSHClient(ctx)
|
||||
require.NoError(t, err)
|
||||
t.Cleanup(func() {
|
||||
|
|
@ -664,6 +668,7 @@ func (c closeFunc) Close() error {
|
|||
func setupAgent(t *testing.T, metadata codersdk.WorkspaceAgentMetadata, ptyTimeout time.Duration) (
|
||||
*codersdk.AgentConn,
|
||||
<-chan *codersdk.AgentStats,
|
||||
afero.Fs,
|
||||
) {
|
||||
if metadata.DERPMap == nil {
|
||||
metadata.DERPMap = tailnettest.RunDERPAndSTUN(t)
|
||||
|
|
@ -671,6 +676,7 @@ func setupAgent(t *testing.T, metadata codersdk.WorkspaceAgentMetadata, ptyTimeo
|
|||
coordinator := tailnet.NewCoordinator()
|
||||
agentID := uuid.New()
|
||||
statsCh := make(chan *codersdk.AgentStats)
|
||||
fs := afero.NewMemMapFs()
|
||||
closer := agent.New(agent.Options{
|
||||
Client: &client{
|
||||
t: t,
|
||||
|
|
@ -679,6 +685,7 @@ func setupAgent(t *testing.T, metadata codersdk.WorkspaceAgentMetadata, ptyTimeo
|
|||
statsChan: statsCh,
|
||||
coordinator: coordinator,
|
||||
},
|
||||
Filesystem: fs,
|
||||
Logger: slogtest.Make(t, nil).Leveled(slog.LevelDebug),
|
||||
ReconnectingPTYTimeout: ptyTimeout,
|
||||
})
|
||||
|
|
@ -704,7 +711,7 @@ func setupAgent(t *testing.T, metadata codersdk.WorkspaceAgentMetadata, ptyTimeo
|
|||
conn.SetNodeCallback(sendNode)
|
||||
return &codersdk.AgentConn{
|
||||
Conn: conn,
|
||||
}, statsCh
|
||||
}, statsCh, fs
|
||||
}
|
||||
|
||||
var dialTestPayload = []byte("dean-was-here123")
|
||||
|
|
|
|||
Loading…
Reference in New Issue