From 6b2e4aea554792bf91d8077238f96d2388511194 Mon Sep 17 00:00:00 2001 From: "denis.bilenko" Date: Wed, 11 Nov 2026 19:54:31 +0000 Subject: [PATCH] acceptance: resolve cloud workspace auth once per process Every cloud acceptance test created its own SDK workspace client, repeating host metadata, OIDC discovery and the token exchange. When the workspace is briefly slow, those requests time out and the default auth chain reports "cannot configure default credentials", failing all tests that start in that window. Share one client across tests; cache only successful results. Co-authored-by: Isaac --- acceptance/internal/prepare_server.go | 51 +++++++++++++++++++++------ 1 file changed, 41 insertions(+), 10 deletions(-) diff --git a/acceptance/internal/prepare_server.go b/acceptance/internal/prepare_server.go index 3890aa52d5a..23ed1118142 100644 --- a/acceptance/internal/prepare_server.go +++ b/acceptance/internal/prepare_server.go @@ -93,6 +93,45 @@ func lookupEnv(testEnv []string, key string) (string, bool) { return "", false } +// cloudWorkspaceState caches the cloud workspace client shared by all tests in +// the process. Resolving auth per test repeats OIDC discovery and the token +// exchange for every test. When the workspace is briefly slow, those requests +// time out, and the SDK's default auth chain hides the timeout behind a +// "cannot configure default credentials" error. Auth resolution is not retried +// the way regular API calls are, so every test running at that moment fails. +// A shared client resolves auth once and refreshes its token in the background. +var cloudWorkspaceState struct { + mu sync.Mutex + client *databricks.WorkspaceClient + user iam.User + workspaceID string +} + +// cloudWorkspace returns the shared workspace client, the current user and the +// workspace id. Only a successful result is cached, so a transient failure +// fails one test and the next test tries again. +func cloudWorkspace(t *testing.T) (*databricks.WorkspaceClient, iam.User, string) { + s := &cloudWorkspaceState + s.mu.Lock() + defer s.mu.Unlock() + + if s.client != nil { + return s.client, s.user, s.workspaceID + } + + w, err := databricks.NewWorkspaceClient() + require.NoError(t, err) + + user, err := iamutil.GetCurrentUser(t.Context(), w) + require.NoError(t, err, "Failed to get current user") + + workspaceID, err := auth.ResolveWorkspaceID(t.Context(), w) + require.NoError(t, err, "Failed to get current workspace id") + + s.client, s.user, s.workspaceID = w, *user, workspaceID + return w, *user, workspaceID +} + func PrepareServerAndClient(t *testing.T, config TestConfig, logRequests bool, outputDir string, testEnv []string) (*sdkconfig.Config, iam.User, string) { cloudEnv := env.Get(t.Context(), "CLOUD_ENV") recordRequests := isTruePtr(config.RecordRequests) @@ -117,15 +156,7 @@ func PrepareServerAndClient(t *testing.T, config TestConfig, logRequests bool, o } if cloudEnv != "" { - w, err := databricks.NewWorkspaceClient() - require.NoError(t, err) - - user, err := iamutil.GetCurrentUser(t.Context(), w) - require.NoError(t, err, "Failed to get current user") - - workspaceID, err := auth.ResolveWorkspaceID(t.Context(), w) - require.NoError(t, err, "Failed to get current workspace id") - + w, user, workspaceID := cloudWorkspace(t) cfg := w.Config // If we are running in a cloud environment AND we need to intercept requests @@ -145,7 +176,7 @@ func PrepareServerAndClient(t *testing.T, config TestConfig, logRequests bool, o } } - return cfg, *user, workspaceID + return cfg, user, workspaceID } // Same topology as cloud, with the testserver as the upstream. Both servers see