feat(webapp): trace deployment outcomes and compute template creation (#4015)

Deploy success/failure wasn't easily observable: compute template
creation only logged, and the terminal deployment statuses (deployed /
failed / timed-out) weren't traced — so deploy health couldn't be seen
without querying the database.

This adds two spans:

- `compute.template.create` around template creation at finalize,
tagging the resolved mode and per-preset outcome. `resolveMode` now
returns its decision (mode + reason) so the span can record why a mode
was chosen.
- `deployment.outcome` via a small shared helper
(`recordDeploymentOutcome`) emitted at every terminal-status write —
finalize (deployed), fail / index-failed / background-worker (failed),
and timeout (timed out) — so deploy success/failure is queryable by
status and reason.

The helper is best-effort (org/project/env enrichment where cheaply
available) and never throws, so telemetry can't break a deploy.
This commit is contained in:
nicktrn
2026-06-22 18:59:59 +01:00
committed by GitHub
parent bb92935c72
commit bf4c6e92bd
7 changed files with 231 additions and 75 deletions
@@ -13,9 +13,26 @@ import { isOrgMigrated } from "~/runEngine/concerns/computeMigration.server";
import { backingForQueue, workerRegionRegistry } from "~/v3/workerRegions.server";
import { globalFlagsRegistry } from "~/v3/globalFlagsRegistry.server";
import { getEntitlement } from "~/services/platform.v3.server";
import { startActiveSpan, attributesFromAuthenticatedEnv } from "~/v3/tracer.server";
type TemplateCreationMode = "required" | "shadow" | "skip";
// Why the mode was chosen — slices the compute.template.create span by path.
type TemplateModeReason =
| "no-client"
| "no-project"
| "microvm-native"
| "migrated"
| "compute-access"
| "rollout"
| "none";
type ResolvedTemplateMode = {
mode: TemplateCreationMode;
migrated: boolean;
reason: TemplateModeReason;
};
type ResolvedPreset = {
name: MachinePresetName;
cpu: number;
@@ -60,89 +77,116 @@ export class ComputeTemplateCreationService {
prisma: PrismaClientOrTransaction;
writer?: WritableStreamDefaultWriter;
}): Promise<void> {
const mode = await this.resolveMode(options.projectId, options.prisma);
return startActiveSpan("compute.template.create", async (span) => {
const { mode, migrated, reason } = await this.resolveMode(
options.projectId,
options.prisma
);
if (mode === "skip") {
return;
}
span.setAttributes({
...attributesFromAuthenticatedEnv(options.authenticatedEnv),
"compute.template.mode": mode,
"compute.template.migrated": migrated,
"compute.template.reason": reason,
"compute.template.deployment_id": options.deploymentFriendlyId,
"compute.template.presets_total": this.presets.length,
"compute.template.presets_required": this.requiredPresets.size,
});
if (mode === "shadow") {
this.createTemplate(options.imageReference, { background: true })
.then((outcome) => {
if (outcome.error) {
logger.error("Shadow template creation failed", {
if (mode === "skip") {
span.setAttribute("compute.template.result", "skipped");
return;
}
if (mode === "shadow") {
// Shadow is fire-and-forget (background build), so the span only records
// that it was dispatched — the build outcome lands server-side later.
span.setAttribute("compute.template.result", "shadow_dispatched");
this.createTemplate(options.imageReference, { background: true })
.then((outcome) => {
if (outcome.error) {
logger.error("Shadow template creation failed", {
id: options.deploymentFriendlyId,
imageReference: options.imageReference,
error: outcome.error,
});
}
})
.catch((error) => {
logger.error("Shadow template creation threw unexpectedly", {
id: options.deploymentFriendlyId,
imageReference: options.imageReference,
error: outcome.error,
error: error instanceof Error ? error.message : String(error),
});
}
})
.catch((error) => {
logger.error("Shadow template creation threw unexpectedly", {
id: options.deploymentFriendlyId,
imageReference: options.imageReference,
error: error instanceof Error ? error.message : String(error),
});
});
return;
}
// Required mode
if (options.writer) {
try {
await options.writer.write(
`event: log\ndata: ${JSON.stringify({ message: "Building compute template..." })}\n\n`
);
} catch {
// Stream may be closed if client disconnected - continue with template creation
return;
}
}
logger.info("Creating compute template (required mode)", {
id: options.deploymentFriendlyId,
imageReference: options.imageReference,
presets: this.presets.map((p) => p.name),
requiredPresets: [...this.requiredPresets],
});
// Required mode
if (options.writer) {
try {
await options.writer.write(
`event: log\ndata: ${JSON.stringify({ message: "Building compute template..." })}\n\n`
);
} catch {
// Stream may be closed if client disconnected - continue with template creation
}
}
const outcome = await this.createTemplate(options.imageReference);
const failureMessage = this.failureMessageForRequiredMode(
outcome,
options.deploymentFriendlyId,
options.imageReference
);
if (failureMessage) {
logger.error("Compute template creation failed", {
logger.info("Creating compute template (required mode)", {
id: options.deploymentFriendlyId,
imageReference: options.imageReference,
error: failureMessage,
presets: this.presets.map((p) => p.name),
requiredPresets: [...this.requiredPresets],
});
const failService = new FailDeploymentService();
await failService.call(options.authenticatedEnv, options.deploymentFriendlyId, {
error: {
name: "TemplateCreationFailed",
message: `Failed to create compute template: ${failureMessage}`,
},
const outcome = await this.createTemplate(options.imageReference);
span.setAttribute("compute.template.presets_built", outcome.results.length);
const failureMessage = this.failureMessageForRequiredMode(
outcome,
options.deploymentFriendlyId,
options.imageReference
);
if (failureMessage) {
span.setAttributes({
"compute.template.result": "failed",
"compute.template.failure": failureMessage,
});
logger.error("Compute template creation failed", {
id: options.deploymentFriendlyId,
imageReference: options.imageReference,
error: failureMessage,
});
const failService = new FailDeploymentService();
await failService.call(options.authenticatedEnv, options.deploymentFriendlyId, {
error: {
name: "TemplateCreationFailed",
message: `Failed to create compute template: ${failureMessage}`,
},
});
throw new ServiceValidationError(`Compute template creation failed: ${failureMessage}`);
}
span.setAttribute("compute.template.result", "created");
logger.info("Compute template created", {
id: options.deploymentFriendlyId,
imageReference: options.imageReference,
results: outcome.results.length,
});
throw new ServiceValidationError(`Compute template creation failed: ${failureMessage}`);
}
logger.info("Compute template created", {
id: options.deploymentFriendlyId,
imageReference: options.imageReference,
results: outcome.results.length,
});
}
async resolveMode(
projectId: string,
prisma: PrismaClientOrTransaction
): Promise<TemplateCreationMode> {
): Promise<ResolvedTemplateMode> {
if (!this.client) {
return "skip";
return { mode: "skip", migrated: false, reason: "no-client" };
}
const project = await prisma.project.findFirst({
@@ -158,11 +202,11 @@ export class ComputeTemplateCreationService {
});
if (!project) {
return "skip";
return { mode: "skip", migrated: false, reason: "no-project" };
}
if (project.defaultWorkerGroup?.workloadType === "MICROVM") {
return "required";
return { mode: "required", migrated: false, reason: "microvm-native" };
}
// Migrated orgs route runs to the compute backing even though their stored
@@ -194,22 +238,26 @@ export class ComputeTemplateCreationService {
}
if (migrated) {
// required => template built at deploy (deploy fails on error); off => shadow.
return decision.flags?.computeMigrationRequireTemplate ? "required" : "shadow";
return {
mode: decision.flags?.computeMigrationRequireTemplate ? "required" : "shadow",
migrated: true,
reason: "migrated",
};
}
}
const hasComputeAccess = await resolveComputeAccess(prisma, project.organization.featureFlags);
if (hasComputeAccess) {
return "shadow";
return { mode: "shadow", migrated: false, reason: "compute-access" };
}
const rolloutPct = Number(env.COMPUTE_TEMPLATE_SHADOW_ROLLOUT_PCT ?? "0");
if (rolloutPct > 0 && Math.random() * 100 < rolloutPct) {
return "shadow";
return { mode: "shadow", migrated: false, reason: "rollout" };
}
return "skip";
return { mode: "skip", migrated: false, reason: "none" };
}
async createTemplate(
@@ -15,6 +15,7 @@ import {
} from "./createBackgroundWorker.server";
import { findOrCreateBackgroundWorker } from "./createDeploymentBackgroundWorkerV4/findOrCreateBackgroundWorker.server";
import { TimeoutDeploymentService } from "./timeoutDeployment.server";
import { recordDeploymentOutcome } from "./recordDeploymentOutcome.server";
import { env } from "~/env.server";
export class CreateDeploymentBackgroundWorkerServiceV4 extends BaseService {
@@ -111,7 +112,7 @@ export class CreateDeploymentBackgroundWorkerServiceV4 extends BaseService {
if (findOrCreateError instanceof ServiceValidationError) {
// `#failBackgroundWorkerDeployment` already throws its argument; the
// outer `throw` covers the non-SVE branch.
await this.#failBackgroundWorkerDeployment(deployment, findOrCreateError);
await this.#failBackgroundWorkerDeployment(deployment, findOrCreateError, environment);
}
throw findOrCreateError;
}
@@ -144,7 +145,7 @@ export class CreateDeploymentBackgroundWorkerServiceV4 extends BaseService {
const serviceError = new ServiceValidationError("Error creating background worker files");
await this.#failBackgroundWorkerDeployment(deployment, serviceError);
await this.#failBackgroundWorkerDeployment(deployment, serviceError, environment);
throw serviceError;
}
@@ -167,7 +168,7 @@ export class CreateDeploymentBackgroundWorkerServiceV4 extends BaseService {
error: resourcesError.message,
});
await this.#failBackgroundWorkerDeployment(deployment, resourcesError);
await this.#failBackgroundWorkerDeployment(deployment, resourcesError, environment);
throw resourcesError;
}
@@ -179,7 +180,7 @@ export class CreateDeploymentBackgroundWorkerServiceV4 extends BaseService {
"Error creating background worker resources"
);
await this.#failBackgroundWorkerDeployment(deployment, serviceError);
await this.#failBackgroundWorkerDeployment(deployment, serviceError, environment);
throw serviceError;
}
@@ -206,7 +207,7 @@ export class CreateDeploymentBackgroundWorkerServiceV4 extends BaseService {
error: schedulesError.message,
});
await this.#failBackgroundWorkerDeployment(deployment, schedulesError);
await this.#failBackgroundWorkerDeployment(deployment, schedulesError, environment);
throw schedulesError;
}
@@ -220,7 +221,7 @@ export class CreateDeploymentBackgroundWorkerServiceV4 extends BaseService {
const serviceError = new ServiceValidationError("Error syncing declarative schedules");
await this.#failBackgroundWorkerDeployment(deployment, serviceError);
await this.#failBackgroundWorkerDeployment(deployment, serviceError, environment);
throw serviceError;
}
@@ -264,7 +265,11 @@ export class CreateDeploymentBackgroundWorkerServiceV4 extends BaseService {
});
}
async #failBackgroundWorkerDeployment(deployment: WorkerDeployment, error: Error) {
async #failBackgroundWorkerDeployment(
deployment: WorkerDeployment,
error: Error,
environment: AuthenticatedEnvironment
) {
// Guarded BUILDING → FAILED transition, symmetric with the BUILDING → DEPLOYING
// transition in `call()`. With idempotent retries, two attempts can run side-by-side;
// without the predicate, one attempt's failure could downgrade the deployment after
@@ -297,6 +302,16 @@ export class CreateDeploymentBackgroundWorkerServiceV4 extends BaseService {
// sibling attempt may have just enqueued it as part of a successful
// BUILDING → DEPLOYING transition.
await TimeoutDeploymentService.dequeue(deployment.id, this._prisma);
recordDeploymentOutcome({
status: "FAILED",
deploymentFriendlyId: deployment.friendlyId,
organizationId: environment.organizationId,
projectId: environment.projectId,
environmentId: environment.id,
environmentType: environment.type,
reason: error.message,
});
}
throw error;
@@ -3,6 +3,7 @@ import { BaseService } from "./baseService.server";
import { logger } from "~/services/logger.server";
import { type WorkerDeploymentStatus } from "@trigger.dev/database";
import { DeploymentService } from "./deployment.server";
import { recordDeploymentOutcome } from "./recordDeploymentOutcome.server";
const FINAL_DEPLOYMENT_STATUSES: WorkerDeploymentStatus[] = [
"CANCELED",
@@ -74,6 +75,16 @@ export class DeploymentIndexFailed extends BaseService {
},
});
recordDeploymentOutcome({
status: "FAILED",
deploymentFriendlyId: deployment.friendlyId,
organizationId: deployment.environment.project.organizationId,
projectId: deployment.environment.projectId,
environmentId: deployment.environmentId,
environmentType: deployment.environment.type,
reason: error.message,
});
const deploymentService = new DeploymentService();
await deploymentService
.appendToEventLog(deployment.environment.project, failedDeployment, [
@@ -5,6 +5,7 @@ import { type WorkerDeploymentStatus } from "@trigger.dev/database";
import { type FailDeploymentRequestBody } from "@trigger.dev/core/v3/schemas";
import { type AuthenticatedEnvironment } from "~/services/apiAuth.server";
import { DeploymentService } from "./deployment.server";
import { recordDeploymentOutcome } from "./recordDeploymentOutcome.server";
export const FINAL_DEPLOYMENT_STATUSES: WorkerDeploymentStatus[] = [
"CANCELED",
@@ -51,6 +52,16 @@ export class FailDeploymentService extends BaseService {
},
});
recordDeploymentOutcome({
status: "FAILED",
deploymentFriendlyId: friendlyId,
organizationId: authenticatedEnv.organizationId,
projectId: authenticatedEnv.projectId,
environmentId: authenticatedEnv.id,
environmentType: authenticatedEnv.type,
reason: params.error.message,
});
const deploymentService = new DeploymentService();
await deploymentService
.appendToEventLog(authenticatedEnv.project, failedDeployment, [
@@ -10,6 +10,7 @@ import { projectPubSub } from "./projectPubSub.server";
import { FailDeploymentService } from "./failDeployment.server";
import { TimeoutDeploymentService } from "./timeoutDeployment.server";
import { DeploymentService } from "./deployment.server";
import { recordDeploymentOutcome } from "./recordDeploymentOutcome.server";
import { engine } from "../runEngine.server";
import { tryCatch } from "@trigger.dev/core";
@@ -78,6 +79,15 @@ export class FinalizeDeploymentService extends BaseService {
},
});
recordDeploymentOutcome({
status: "DEPLOYED",
deploymentFriendlyId: deployment.friendlyId,
organizationId: authenticatedEnv.organizationId,
projectId: authenticatedEnv.projectId,
environmentId: authenticatedEnv.id,
environmentType: authenticatedEnv.type,
});
const deploymentService = new DeploymentService();
await deploymentService
.appendToEventLog(authenticatedEnv.project, finalizedDeployment, [
@@ -0,0 +1,50 @@
import { SpanStatusCode } from "@opentelemetry/api";
import { type WorkerDeploymentStatus } from "@trigger.dev/database";
import { logger } from "~/services/logger.server";
import { tracer } from "~/v3/tracer.server";
type TerminalDeploymentStatus = Extract<
WorkerDeploymentStatus,
"DEPLOYED" | "FAILED" | "TIMED_OUT"
>;
/**
* Records a deployment's terminal status as a `deployment.outcome` span so
* deploy success/failure is queryable from traces (no DB read). Call after each
* terminal-status write. Org/project/env are best-effort; never throws.
*/
export function recordDeploymentOutcome(params: {
status: TerminalDeploymentStatus;
deploymentFriendlyId: string;
organizationId?: string;
projectId?: string;
environmentId?: string;
environmentType?: string;
reason?: string;
}): void {
try {
const span = tracer.startSpan("deployment.outcome", {
attributes: {
"$trigger.org.id": params.organizationId,
"$trigger.project.id": params.projectId,
"$trigger.env.id": params.environmentId,
"$trigger.env.type": params.environmentType,
"deployment.outcome.status": params.status,
"deployment.outcome.success": params.status === "DEPLOYED",
"deployment.outcome.deployment_id": params.deploymentFriendlyId,
"deployment.outcome.reason": params.reason,
},
});
if (params.status !== "DEPLOYED") {
span.setStatus({ code: SpanStatusCode.ERROR, message: params.reason });
}
span.end();
} catch (error) {
logger.debug("recordDeploymentOutcome failed", {
deploymentFriendlyId: params.deploymentFriendlyId,
error: error instanceof Error ? error.message : String(error),
});
}
}
@@ -5,6 +5,7 @@ import { PerformDeploymentAlertsService } from "./alerts/performDeploymentAlerts
import { type PrismaClientOrTransaction } from "~/db.server";
import { workerQueue } from "~/services/worker.server";
import { DeploymentService } from "./deployment.server";
import { recordDeploymentOutcome } from "./recordDeploymentOutcome.server";
export class TimeoutDeploymentService extends BaseService {
public async call(id: string, fromStatus: string, errorMessage: string) {
@@ -48,6 +49,16 @@ export class TimeoutDeploymentService extends BaseService {
},
});
recordDeploymentOutcome({
status: "TIMED_OUT",
deploymentFriendlyId: deployment.friendlyId,
organizationId: deployment.environment.project.organizationId,
projectId: deployment.environment.projectId,
environmentId: deployment.environmentId,
environmentType: deployment.environment.type,
reason: errorMessage,
});
const deploymentService = new DeploymentService();
await deploymentService
.appendToEventLog(deployment.environment.project, timedOutDeployment, [