Cloud editing MVP

admin/jobs/river/reconcile_deployment.go

@@ -3,6 +3,8 @@ package river
 import (
 	"context"
 	"errors"
+	"strings"
+	"time"
 
 	"github.com/jackc/pgx/v5"
 	"github.com/rilldata/rill/admin"
@@ -24,6 +26,12 @@ type ReconcileDeploymentWorker struct {
 	admin *admin.Service
 }
 
+// NextRetryAt uses exponential backoff starting at 15s: ~15s, ~30s, ~60s, ~120s.
+// This keeps total retry duration under ~4 minutes so users aren't stuck waiting.
+func (w *ReconcileDeploymentWorker) NextRetryAt(job *river.Job[ReconcileDeploymentArgs]) time.Time {
+	return time.Now().Add(15 * time.Second * time.Duration(1<<(job.Attempt-1)))
+}
+
 // NewReconcileDeploymentWorker creates a new ReconcileDeploymentWorker. Only to be used in tests to trigger the worker directly.
 func NewReconcileDeploymentWorker(admin *admin.Service) *ReconcileDeploymentWorker {
 	return &ReconcileDeploymentWorker{
@@ -74,8 +82,14 @@ func (w *ReconcileDeploymentWorker) Work(ctx context.Context, job *river.Job[Rec
 			}
 
 			// Initialize the deployment (by provisioning a runtime and creating an instance on it)
-			err := w.admin.StartDeploymentInner(ctx, depl)
-			if err != nil {
+			if err := w.admin.StartDeploymentInner(ctx, depl); err != nil {
+				// Check if this is a non-retryable error (fail fast instead of waiting for all retries)
+				if isNonRetryable(err) || job.Attempt >= job.MaxAttempts {
+					if _, dbErr := w.admin.DB.UpdateDeploymentStatus(ctx, depl.ID, database.DeploymentStatusErrored, err.Error()); dbErr != nil {
+						w.admin.Logger.Error("reconcile deployment: failed to set errored status", observability.ZapCtx(ctx), zap.Error(dbErr))
+					}
+					return river.JobCancel(err)
+				}
 				return err
 			}
 		}
@@ -115,12 +129,24 @@ func (w *ReconcileDeploymentWorker) Work(ctx context.Context, job *river.Job[Rec
 		// Delete the deployment and all its resources.
 		err := w.admin.StopDeploymentInner(ctx, depl)
 		if err != nil {
+			if job.Attempt >= job.MaxAttempts {
+				if _, dbErr := w.admin.DB.UpdateDeploymentStatus(ctx, depl.ID, database.DeploymentStatusErrored, err.Error()); dbErr != nil {
+					w.admin.Logger.Error("reconcile deployment: failed to set errored status during deletion", observability.ZapCtx(ctx), zap.Error(dbErr))
+				}
+				return river.JobCancel(err)
+			}
 			return err
 		}
 
 		// Delete the deployment
 		err = w.admin.DB.DeleteDeployment(ctx, depl.ID)
 		if err != nil {
+			if job.Attempt >= job.MaxAttempts {
+				if _, dbErr := w.admin.DB.UpdateDeploymentStatus(ctx, depl.ID, database.DeploymentStatusErrored, err.Error()); dbErr != nil {
+					w.admin.Logger.Error("reconcile deployment: failed to set errored status during deletion", observability.ZapCtx(ctx), zap.Error(dbErr))
+				}
+				return river.JobCancel(err)
+			}
 			return err
 		}
 
@@ -153,3 +179,12 @@ func (w *ReconcileDeploymentWorker) Work(ctx context.Context, job *river.Job[Rec
 
 	return nil
 }
+
+// isNonRetryable returns true for errors that won't resolve with retries,
+// such as capacity limits or configuration errors.
+func isNonRetryable(err error) bool {
+	msg := err.Error()
+	return strings.Contains(msg, "no runtimes found with sufficient available slots") ||
+		strings.Contains(msg, "Invalid environment") ||
+		strings.Contains(msg, "not a valid version")
+}

admin/jobs/river/river.go

@@ -335,7 +335,7 @@ func (c *Client) ReconcileDeployment(ctx context.Context, deploymentID string) (
 	res, err := c.riverClient.Insert(ctx, ReconcileDeploymentArgs{
 		DeploymentID: deploymentID,
 	}, &river.InsertOpts{
-		MaxAttempts: 25, // Last retry, ~3 weeks after first run
+		MaxAttempts: 5, // Retries at ~15s, 30s, 60s, 120s (see NextRetryAt override)
 		UniqueOpts: river.UniqueOpts{
 			ByArgs: true,
 			ByState: []rivertype.JobState{

admin/server/deployment.go

@@ -241,6 +241,7 @@
 	}
 	if depl.Environment == "dev" {
 		instancePermissions = append(instancePermissions,
+			runtime.ReadInstance,
 			runtime.ReadOLAP,
 			runtime.ReadProfiling,
 			runtime.ReadRepo,
@@ -1031,3 +1032,4 @@
 
 	return attr, userID, forProjPerms.ReadProd, nil
 }
+

admin/server/projects.go

@@ -31,7 +31,7 @@ import (
 
 const devDeplTTL = 6 * time.Hour
 
-const devSlots = 8
+const devSlots = 4
 
 const prodDeplTTL = 14 * 24 * time.Hour
 
@@ -456,6 +456,13 @@ func (s *Server) GetProject(ctx context.Context, req *adminv1.GetProjectRequest)
 			runtime.EditTrigger,
 		)
 	}
+	if req.Branch != "" && permissions.ManageDev {
+		instancePermissions = append(
+			instancePermissions,
+			runtime.ReadRepo,
+			runtime.EditRepo,
+		)
+	}
 
 	var systemPermissions []runtime.Permission
 	if req.IssueSuperuserToken {
@@ -2178,6 +2185,7 @@ func (s *Server) projToDTO(p *database.Project, orgName string) *adminv1.Project
 		Provisioner:         p.Provisioner,
 		ProdVersion:         p.ProdVersion,
 		ProdSlots:           int64(p.ProdSlots),
+		DevSlots:            int64(p.DevSlots),
 		PrimaryBranch:       p.PrimaryBranch,
 		Subpath:             p.Subpath,
 		GitRemote:           safeStr(p.GitRemote),