fix(deploy): keep the healthy new container when a rolling update fails late

Laravel runs failed() on a fresh copy of the job, so the in-memory
newVersionIsHealthy flag was always false there and a failure after the
old container stopped removed the only running container. Keep the flag
in the cache, keyed by the deployment UUID.

Closes #12253

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
This commit is contained in:
Andras Bacsai
2026-10-23 23:03:33 +02:00
co-authored by Claude Opus 5.5
parent 2a7faa1210
commit 1bebdffaff
4 changed files with 131 additions and 22 deletions
@@ -6,6 +6,9 @@ use App\Models\Application;
use App\Models\ApplicationDeploymentQueue;
use Illuminate\Contracts\Queue\Job;
use Illuminate\Queue\TimeoutExceededException;
use Tests\TestCase;
uses(TestCase::class);
/**
* Test to verify that deployment errors are properly logged with comprehensive details.
@@ -91,6 +94,7 @@ it('logs comprehensive error details when failed() is called', function () {
$containerNameProperty = $reflection->getProperty('container_name');
$containerNameProperty->setAccessible(true);
$containerNameProperty->setValue($job, 'test-container');
$reflection->getProperty('deployment_uuid')->setValue($job, 'error-logging-test');
// Mock the failDeployment method to prevent errors
$job->shouldReceive('failDeployment')->andReturn();
@@ -197,6 +201,7 @@ it('handles exceptions with no message gracefully', function () {
$containerNameProperty = $reflection->getProperty('container_name');
$containerNameProperty->setAccessible(true);
$containerNameProperty->setValue($job, 'test-container');
$reflection->getProperty('deployment_uuid')->setValue($job, 'error-logging-test');
$job->shouldReceive('failDeployment')->andReturn();
$job->shouldReceive('execute_remote_command')->andReturn();
@@ -288,6 +293,7 @@ it('logs error code 0 correctly', function () {
$containerNameProperty = $reflection->getProperty('container_name');
$containerNameProperty->setAccessible(true);
$containerNameProperty->setValue($job, 'test-container');
$reflection->getProperty('deployment_uuid')->setValue($job, 'error-logging-test');
$job->shouldReceive('failDeployment')->andReturn();
$job->shouldReceive('execute_remote_command')->andReturn();
@@ -393,6 +399,7 @@ it('logs the configured timeout when the deployment times out', function () {
$containerNameProperty = $reflection->getProperty('container_name');
$containerNameProperty->setValue($job, 'test-container');
$reflection->getProperty('deployment_uuid')->setValue($job, 'error-logging-test');
$job->shouldReceive('failDeployment')->andReturn();
$job->shouldReceive('execute_remote_command')->andReturn();
@@ -4,11 +4,14 @@ use App\Jobs\ApplicationDeploymentJob;
use App\Models\Application;
use App\Models\ApplicationDeploymentQueue;
use App\Models\ApplicationSetting;
use Tests\TestCase;
uses(TestCase::class);
/**
* @return array<int, array<int|string, mixed>>
*/
function runFailedDeploymentCleanup(bool $consistentContainerName, ?string $customInternalName, int $pullRequestId, bool $newVersionIsHealthy = false): array
function runFailedDeploymentCleanup(bool $consistentContainerName, ?string $customInternalName, int $pullRequestId): array
{
$settings = new ApplicationSetting;
$settings->setRawAttributes([
@@ -36,8 +39,8 @@ function runFailedDeploymentCleanup(bool $consistentContainerName, ?string $cust
$reflection->getProperty('application_deployment_queue')->setValue($job, $queue);
$reflection->getProperty('application')->setValue($job, $application);
$reflection->getProperty('pull_request_id')->setValue($job, $pullRequestId);
$reflection->getProperty('deployment_uuid')->setValue($job, 'cleanup-test-deployment');
$reflection->getProperty('container_name')->setValue($job, 'new-container-20261001T120000');
$reflection->getProperty('newVersionIsHealthy')->setValue($job, $newVersionIsHealthy);
$job->failed(new RuntimeException('Health check failed'));
@@ -62,14 +65,3 @@ it('keeps the running container when consistent container names are enabled', fu
it('keeps the running container for pull request deployments', function () {
expect(runFailedDeploymentCleanup(consistentContainerName: false, customInternalName: 'my-api', pullRequestId: 12))->toBeEmpty();
});
it('keeps the new container when the deployment fails after the new container became healthy', function () {
expect(runFailedDeploymentCleanup(consistentContainerName: false, customInternalName: null, pullRequestId: 0, newVersionIsHealthy: true))->toBeEmpty();
});
it('removes the new container when the deployment fails before the new container became healthy', function () {
$remoteCommands = runFailedDeploymentCleanup(consistentContainerName: false, customInternalName: null, pullRequestId: 0, newVersionIsHealthy: false);
expect($remoteCommands)->toHaveCount(1)
->and($remoteCommands[0][0])->toContain('new-container-20261001T120000');
});