https://pulumi.com logo
Join Slack
Powered by
This message was deleted.
# general
s
This message was deleted.
b
Hi @bitter-river-54082 Please can you show me the code that is causing you the issues here?
b
This is the code of my 
migration-provisioner
that is executing before the NAT gateway created :
Copy code
export const Migration = new Provisioner<string, never>("migration-provisioner", {
    dep: config.version.version,
    onCreate: runMigration,
    changeToken: "",
}, {
    dependsOn: [
        ec2.servicesInstance,
        ec2.servicesDataVolume,
        ssm.runMigrationDocument,
    ],
})

export const delay = (t: number) => new Promise(resolve => setTimeout(resolve, t))

function runMigration(): Promise<never> {
    return new Promise((resolve, reject) => {
        const runMigrationDocumentName = ssm.runMigrationDocument.name.get()
        const servicesInstanceId = ec2.servicesInstance.id.get()
        const ssmClient = new AWS.SSM({region: config.aws.region})

        ssmClient.sendCommand({
            DocumentName: runMigrationDocumentName,
            InstanceIds: [servicesInstanceId],
            Parameters: {
                "dryRun": ["false"],
            },
        }).promise().then(response => delay(3000).then(() => ssmClient.waitFor("commandExecuted", {
            CommandId: response.Command!.CommandId!,
            InstanceId: servicesInstanceId,
            $waiter: {
                delay: 5,
                maxAttempts: 60,
            },
        }).promise().then(response => {
            if (response.Status != "Success") {
                throw new Error(`Migration failed with status '${response.Status}'`)
            }
            resolve()
        }))).catch(reject)
    })
}
b
Ok, I need to see the program where you are defining your VPC + Instance
b
Copy code
export const vpc = new awsx.ec2.Vpc("vpc", {
    cidrBlock: config.vpc?.cidrBlock,
    numberOfAvailabilityZones: 2,
    tags: {
        deployment: config.deploymentName,
    },
})
b
ok, so you are then deploying instances right? Where do you relate the instance with the VPC?
b
Instance :
Copy code
export const servicesInstance = new aws.ec2.Instance("services", {
    ami: config.ami.services[config.aws.region],
    ebsOptimized: true,
    ebsBlockDevices: [{
        deviceName: "/dev/sdf",
        encrypted: true,
        volumeSize: 100,
        volumeType: "gp2",
    }],
    iamInstanceProfile: iam.servicesInstanceProfile,
    instanceType: aws.ec2.InstanceTypes.M5_Large,
    keyName: config.aws.keyPair,
    metadataOptions: {
        httpEndpoint: "enabled",
        httpPutResponseHopLimit: 2,
        httpTokens: "required",
    },
    rootBlockDevice: {
        encrypted: true,
    },
    subnetId: pulumi.output(vpc.vpc.privateSubnetIds).apply(subnets => subnets[0]),
    tags: {
        deployment: config.deploymentName,
        role: "services",
    },
    userData: pulumi.all([
        s3.configBucket.bucket,
        efs.capsuleCache?.id,
        efs.datasets.id,
        config.stackname,
    ]).apply(([
        configBucketName,
        capsuleCacheEfsId,
        datasetsEfsId,
        pulumiStackName,
    ]) => {
        const template = handlebars.compile(fs.readFileSync("ec2-init-services.sh", "utf8"))
        return template({
            configBucketName,
            capsuleCacheEfsId,
            datasetsEfsId,
            pulumiStackName,
        })
    }),
    vpcSecurityGroupIds: [vpc.sgServices.id],
}, {
    dependsOn: [
        vpc.vpc,
        efs.capsuleCache,
        efs.datasets,
        s3.configBucket,
        cloudwatch.instancesLogGroup,
        cloudwatch.servicesLogGroup,
    ],
    ignoreChanges: ["ebsBlockDevices"],
})
b
so you have this in your instance:
Copy code
subnetId: pulumi.output(vpc.vpc.privateSubnetIds).apply(subnets => subnets[0]),
So if the instance isn't waiting on the VPC then this would fail
So it means that the Migration is failing to wait for the Instance to be fully available right?
b
yes
b
ok, so your Migration code is going to have to do the work of checking to see if the instance is "REady" - once a successful creation of an instance is achieved from the API, then Pulumi's job is effectively over I'm afraid...
b
ok, how should it be done?
f
@broad-dog-22463 the issue that we have is that the EC2 instance is brought up before all VPC resources have been provisioned. Specifically, it is created before the NAT gateways. There is no explicit dependency between the ec2 instance resource and the NAT gateway resources, but there is an implicit dependency since the instance relies on internet access to perform its cloud init scripts. We hoped that adding a dependency on the 
awsx.ec2.Vpc
will do the trick but we don't see the desired effect.
2 Views
Open in Slack
PreviousNext
Powered by