Issue using Terraform EC2 Userdata

Question

I am deploying a bunch of EC2 instances that require a mount called /data, this is a seperate disk that I am attaching using volume attach in AWS.

Now when I did the following manually it works fine, so the script I use works however when adding it via userdata I am seeing issues and the mkfs command is not happening.

If you see my terraform config:

resource "aws_instance" "riak" {
  count = 5
  ami = "${var.aws_ami}"
  vpc_security_group_ids = ["${aws_security_group.bastion01_sg.id}","${aws_security_group.riak_sg.id}","${aws_security_group.outbound_access_sg.id}"]
  subnet_id = "${element(module.vpc.database_subnets, 0)}"
  instance_type = "m4.xlarge"
  tags {
    Name = "x_riak_${count.index}"
    Role = "riak"
  }
  root_block_device {
    volume_size = 20
  }
  user_data = "${file("datapartition.sh")}"
}

resource "aws_volume_attachment" "riak_data" {
  count = 5
  device_name = "/dev/sdh"
  volume_id  = "${element(aws_ebs_volume.riak_data.*.id, count.index)}"
  instance_id = "${element(aws_instance.riak.*.id, count.index)}"
}

And then the partition script is as follows:

#!/bin/bash

if [ ! -d /data ];
then mkdir /data
fi

/sbin/mkfs -t ext4 /dev/xvdh;

while [ -e /dev/xvdh ] ; do sleep 1 ; done

mount /dev/xvdh /data

echo "/dev/xvdh /data ext4 defaults 0 2" >> /etc/fstab

Now when I do this via terraform the mkfs doesn't appear to happen and I see no obvious errors in the syslog. If I copy the script manually and just bash script.sh the mount is created and works as expected.

Has anyone got any suggestions here?

Edit: It's wort noting adding this in AWS gui under userdata also works fine.

Likely /dev/xvdh hasn't been attached yet; why don't you throw a 30 second sleep after that first conditional and give it a try — TJ Biddle
– TJ Biddle, Commented Dec 8, 2017 at 13:51
Thanks for this, originally there was a 30 second sleep it must have been lost in an edit. I also thought it was this, adding log lines suggested otherwise (I could see all my log lines but not any to do with the mkfs). Thanks for the response — Anthony 'Runt' Cleaves
– Anthony 'Runt' Cleaves, Commented Dec 8, 2017 at 14:09

jseguillon · Accepted Answer · 2017-12-08 13:58:39Z

3

You could try with remote_exec instead of user_data.

User_data relates on cloud-init which can act differently depending on images of your cloud provider.

And also i'm not sure it's a good idea to exec a script that would wait for some time before executing in the cloud-init section => this may lead to VM considering launch has failed because of a timeout (depending on your cloud provider).

Remote_exec may be better here because you will be able to wait until your /dev/xvdh is attached

See here

resource "aws_instance" "web" {
  # ...

  provisioner "file" {
    source      = "script.sh"
    destination = "/tmp/script.sh"
  }

  provisioner "remote-exec" {
    inline = [
      "chmod +x /tmp/script.sh",
      "/tmp/script.sh args",
    ]
  }
}

answered Dec 8, 2017 at 13:58

jseguillon

4132 silver badges10 bronze badges

Sign up to request clarification or add additional context in comments.

7 Comments

jseguillon Over a year ago

And also here : github.com/hashicorp/terraform/issues/4668 => some hacks regarding how to wait for full boot before launching a script

Anthony 'Runt' Cleaves Over a year ago

I have a feeling that the volume attachment is happening after this script is being run. This caused timeouts with the volumes being created.

Anthony 'Runt' Cleaves Over a year ago

I have moved it to the aws_volume_attachment block and re-testing

Anthony 'Runt' Cleaves Over a year ago

Same issue :( ` aws_volume_attachment.riak_data.0: Still creating... (3m40s elapsed) aws_volume_attachment.riak_data.1: Still creating... (3m50s elapsed) aws_volume_attachment.riak_data.3: Still creating... (3m50s elapsed) aws_volume_attachment.riak_data.2: Still creating... (3m50s elapsed) aws_volume_attachment.riak_data.4: Still creating... (3m50s elapsed) `

Anthony 'Runt' Cleaves Over a year ago

I certainly can, I believe this may be an issue with that fact I use a bastion host though.

|

Collectives™ on Stack Overflow

Issue using Terraform EC2 Userdata

1 Answer 1

7 Comments

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

1 Answer 1

7 Comments

Your Answer

Sign up or log in

Post as a guest

Related