New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

EC2 instances with EBS volumes destroyed & recreated on each apply #5006

Closed
bwhaley opened this Issue Feb 5, 2016 · 17 comments

Comments

Projects
None yet
@bwhaley

bwhaley commented Feb 5, 2016

Running v0.6.11, I noticed that instances with an ebs_block_device are recreated every time I run terraform apply, even if there were no relevant changes. The block looks like this:

ebs_block_device {
    device_name  = "${var.es_ebs_vol.device_name}"
    volume_type = "${var.es_ebs_vol.type}"
    volume_size = "${var.es_ebs_vol.size}"
}

This seems similar to #913 but that was resolved some time ago. Any ideas?

@tiyberius

This comment has been minimized.

tiyberius commented Feb 6, 2016

+1

Having the same issue as @bwhaley on 0.6.11. If it helps, I had the issue on 0.6.9 as well. I was hoping that an upgrade to 0.6.11 would fix it but it has not :(

@davedash

This comment has been minimized.

Contributor

davedash commented Feb 26, 2016

I'm trying to figure this out too in #4786.

Does your AMI specify an EBS snapshot to mount as a root device? This is my problem with trying to launch and ECS container.

@bwhaley

This comment has been minimized.

bwhaley commented Feb 27, 2016

Yes - but isn't the root volume for all EBS-backed AMIs started from an EBS snapshot?

@davedash

This comment has been minimized.

Contributor

davedash commented Feb 29, 2016

@bwhaley I guess that's the case.

So this is only happening on my t2 based instances that happen to be running Amazon Linux. This might just be a coincidence.

@bwhaley

This comment has been minimized.

bwhaley commented Feb 29, 2016

Interesting - in my case it's also Amazon Linux on T2 instance types.

@octalthorpe

This comment has been minimized.

octalthorpe commented Mar 7, 2016

***** Forget that need to use "root_block_device" my mistake
+1
Seeing the same issue here across more than just t2 instances.

ebs_block_device {
    device_name = "/dev/xvda"
    volume_size = 128
  }

Same behaviour on 0.6.9 thru 0.6.12, possible change on the AWS side?

@eedwardsdisco

This comment has been minimized.

eedwardsdisco commented Mar 9, 2016

+1 I'm hitting this with 0.6.12

Launching an AMI that's a t2.micro with 2 EBS volumes, created using Packer

I specify a "root_block_device" and "data_block_device" mapping in the terraform template.

It's causing it to mark the data volume as needing to be re-created every time.

@Sreeramk

This comment has been minimized.

Sreeramk commented May 11, 2016

I am running into a similar issue. In the plan it keeps pointing to the following.
Couple of observations: I am not using iops. Yet it seems to compute iops
I did not change the delete_on_terminate - yet it thinks there is a change
ebs_block_device.3796809015.delete_on_termination:* "true" => "1"* (forces new resource)
ebs_block_device.3796809015.device_name: "/dev/xvdl" => "/dev/xvdl" (forces new resource)
ebs_block_device.3796809015.encrypted: "false" => ""
ebs_block_device.3796809015.iops: "150" => ""
ebs_block_device.3796809015.snapshot_id: "" => "" (forces new resource)
ebs_block_device.3796809015.volume_size: "50" => "50" (forces new resource)
ebs_block_device.3796809015.volume_type: "gp2" => "gp2" (forces new resource)

@serdardalgic

This comment has been minimized.

serdardalgic commented Jul 18, 2016

I'm also hitting the same issue here:

> terraform plan output
...
    ami:                                               "ami-d3a04fbc" => "ami-d3a04fbc"
    associate_public_ip_address:                       "false" => "false"
    availability_zone:                                 "eu-central-1a" => "<computed>"
    ebs_block_device.#:                                "3" => "3"
    ebs_block_device.1376874904.delete_on_termination: "true" => "false"
    ebs_block_device.1376874904.device_name:           "/dev/xvdf" => ""
    ebs_block_device.1494882292.delete_on_termination: "true" => "false"
    ebs_block_device.1494882292.device_name:           "/dev/xvdh" => ""
    ebs_block_device.1712666200.delete_on_termination: "true" => "false"
    ebs_block_device.1712666200.device_name:           "/dev/xvdg" => ""
    ebs_block_device.3846643179.delete_on_termination: "" => "true" (forces new resource)
    ebs_block_device.3846643179.device_name:           "" => "/dev/xvdh" (forces new resource)
    ebs_block_device.3846643179.encrypted:             "" => "<computed>" (forces new resource)
    ebs_block_device.3846643179.iops:                  "" => "100" (forces new resource)
    ebs_block_device.3846643179.snapshot_id:           "" => "<computed>" (forces new resource)
    ebs_block_device.3846643179.volume_size:           "" => "10" (forces new resource)
    ebs_block_device.3846643179.volume_type:           "" => "io1" (forces new resource)
    ebs_block_device.3994770134.delete_on_termination: "" => "true" (forces new resource)
    ebs_block_device.3994770134.device_name:           "" => "/dev/xvdg" (forces new resource)
    ebs_block_device.3994770134.encrypted:             "" => "<computed>" (forces new resource)
    ebs_block_device.3994770134.iops:                  "" => "250" (forces new resource)
    ebs_block_device.3994770134.snapshot_id:           "" => "<computed>" (forces new resource)
    ebs_block_device.3994770134.volume_size:           "" => "25" (forces new resource)
    ebs_block_device.3994770134.volume_type:           "" => "io1" (forces new resource)
    ebs_block_device.4023988449.delete_on_termination: "" => "true" (forces new resource)
    ebs_block_device.4023988449.device_name:           "" => "/dev/xvdf" (forces new resource)
    ebs_block_device.4023988449.encrypted:             "" => "<computed>" (forces new resource)
    ebs_block_device.4023988449.iops:                  "" => "2000" (forces new resource)
    ebs_block_device.4023988449.snapshot_id:           "" => "<computed>" (forces new resource)
    ebs_block_device.4023988449.volume_size:           "" => "800" (forces new resource)
    ebs_block_device.4023988449.volume_type:           "" => "io1" (forces new resource)
...
    root_block_device.#:                               "1" => "<computed>"
    security_groups.#:                                 "0" => "<computed>"

Although it's already in the terraform.tfstate

                "aws_instance.mongod": {
                    "type": "aws_instance",
                    "primary": {
                        "id": "i-32f24a8f",
                        "attributes": {
                            "ami": "ami-d3a04fbc",
                            "associate_public_ip_address": "false",
                            "availability_zone": "eu-central-1a",
                            "disable_api_termination": "false",
                            "ebs_block_device.#": "3",
                            "ebs_block_device.1376874904.delete_on_termination": "true",
                            "ebs_block_device.1376874904.device_name": "/dev/xvdf",
                            "ebs_block_device.1376874904.encrypted": "false",
                            "ebs_block_device.1376874904.iops": "2000",
                            "ebs_block_device.1376874904.snapshot_id": "snap-e726c30c",
                            "ebs_block_device.1376874904.volume_size": "800",
                            "ebs_block_device.1376874904.volume_type": "io1",
                            "ebs_block_device.1494882292.delete_on_termination": "true",
                            "ebs_block_device.1494882292.device_name": "/dev/xvdh",
                            "ebs_block_device.1494882292.encrypted": "false",
                            "ebs_block_device.1494882292.iops": "100",
                            "ebs_block_device.1494882292.snapshot_id": "snap-398501d2",
                            "ebs_block_device.1494882292.volume_size": "10",
                            "ebs_block_device.1494882292.volume_type": "io1",
                            "ebs_block_device.1712666200.delete_on_termination": "true",
                            "ebs_block_device.1712666200.device_name": "/dev/xvdg",
                            "ebs_block_device.1712666200.encrypted": "false",
                            "ebs_block_device.1712666200.iops": "250",
                            "ebs_block_device.1712666200.snapshot_id": "snap-b60e035e",
                            "ebs_block_device.1712666200.volume_size": "25",
                            "ebs_block_device.1712666200.volume_type": "io1",
                            "ebs_optimized": "false",
                            "ephemeral_block_device.#": "0",
...

Please tell me if you need more info about the issue.

@serdardalgic

This comment has been minimized.

serdardalgic commented Jul 19, 2016

I've solved my problem, just to inform you: I was building with AMIs that already have 3 ebs block devices. Then, in terraform, I was provisioning them with cloud-init. That's why in terraform plan output, there are actually 6 different ebs_block_device ids that were creating the trouble. So, on my side, the problem does not exist. Sorry for the confusion.

@madamedwards

This comment has been minimized.

madamedwards commented Jul 29, 2016

I'm having a similar issue. Mine seems to have something to do with the volumes being encrypted. When I run terraform apply, I get the following:

kms_key_id: "arn:aws:kms:us-east-1:<account_id>:key/<key_id>" => "<key_id (unchanged)>" (forces new resource)

As noted, the key did not change.

@dennybaa

This comment has been minimized.

dennybaa commented Aug 5, 2016

+1

@mr510

This comment has been minimized.

mr510 commented Aug 15, 2016

+1 In terraform v 0.7 I am getting similar error. When encrypting devices in a aws_db_instance.

kms_key_id: "arn:aws:kms:us-west-2::key/key_id" => "Key-ID" (forces new resource)

Key id is the same and have not changed.

@mr510

This comment has been minimized.

mr510 commented Aug 15, 2016

Hmm so the issues can be resolved by using the arn instead of key_id

Forces new resource kms_key_id = "${aws_kms_key.key_name.key_id}"

Change it to kms_key_id = "${aws_kms_key.key_name.arn}". No new resource created.

@iwat

This comment has been minimized.

iwat commented Nov 9, 2016

Same here, terraform tried to re-create EC2 instance with additional EBS.

ebs_block_device.#:                                "1" => "1"
ebs_block_device.1399095401.delete_on_termination: "true" => "false"
ebs_block_device.1399095401.device_name:           "/dev/sdb" => ""
ebs_block_device.2576023345.delete_on_termination: "" => "true" (forces new resource)
ebs_block_device.2576023345.device_name:           "" => "/dev/sdb" (forces new resource)
ebs_block_device.2576023345.encrypted:             "" => "<computed>" (forces new resource)
ebs_block_device.2576023345.iops:                  "" => "<computed>" (forces new resource)
ebs_block_device.2576023345.snapshot_id:           "" => "<computed>" (forces new resource)
ebs_block_device.2576023345.volume_size:           "" => "8" (forces new resource)
ebs_block_device.2576023345.volume_type:           "" => "gp2" (forces new resource)

Inside terraform.tfstate:

"ebs_block_device.#": "1",
"ebs_block_device.1399095401.delete_on_termination": "true",
"ebs_block_device.1399095401.device_name": "/dev/sdb",
"ebs_block_device.1399095401.encrypted": "false",
"ebs_block_device.1399095401.iops": "100",
"ebs_block_device.1399095401.snapshot_id": "snap-9c8bf919",
"ebs_block_device.1399095401.volume_size": "8",
"ebs_block_device.1399095401.volume_type": "gp2",

It stops re-creating instance if I change my .tf file from:

ebs_block_device {
    device_name           = "/dev/sdb"
    volume_type           = "gp2"
    volume_size           = 8
    delete_on_termination = true
}

to

ebs_block_device {
    device_name           = "/dev/sdb"
    volume_type           = "gp2"
    volume_size           = 8
    delete_on_termination = true
    snapshot_id           = "snap-9c8bf919"
}
@jurajseffer

This comment has been minimized.

jurajseffer commented Nov 23, 2016

terraform 0.7.11 here. I use remote statefile (S3). When I run plan after apply, terraform reports:

ebs_block_device.#:                                "0" => "1"
ebs_block_device.3239300295.delete_on_termination: "" => "false" (forces new resource)
ebs_block_device.3239300295.device_name:           "" => "/dev/sda1" (forces new resource)
ebs_block_device.3239300295.encrypted:             "" => "<computed>" (forces new resource)
ebs_block_device.3239300295.iops:                  "" => "<computed>" (forces new resource)
ebs_block_device.3239300295.snapshot_id:           "" => "<computed>" (forces new resource)
ebs_block_device.3239300295.volume_size:           "" => "100" (forces new resource)
ebs_block_device.3239300295.volume_type:           "" => "gp2" (forces new resource)

Statefile has the following for the instance:

"root_block_device.#": "1",
"root_block_device.0.delete_on_termination": "false",
"root_block_device.0.iops": "300",
"root_block_device.0.volume_size": "100",
"root_block_device.0.volume_type": "gp2",

config

ebs_block_device {
    device_name = "/dev/sda1"
    volume_type = "gp2"
    volume_size = "${var.instance_volume_size}"
    delete_on_termination = false
  }
@jstlaurent

This comment has been minimized.

jstlaurent commented Dec 9, 2016

I'm also getting this with Terraform 0.7.13.

My instance definition looks like this:

resource "aws_instance" "app_instance" {
  ami = "${data.aws_ami.ecs_optimized.id}"
  instance_type = "${var.app_instance["instance_type"]}"
  count = "${var.app_instance["instance_count"]}"

  # Some storage
  ebs_block_device {
    device_name = "/dev/sdb"
    volume_size = 50
    volume_type = "gp2"
  }
  ebs_block_device {
    device_name = "/dev/sdc"
    volume_size = 50
    volume_type = "gp2"
  }
  ebs_block_device {
    device_name = "/dev/sdd"
    volume_size = 50
    volume_type = "gp2"
  }
  ebs_block_device {
    device_name = "/dev/sde"
    volume_size = 50
    volume_type = "gp2"
  }
}

I get this every time I run plan:

ebs_block_device.#:                                "5" => "4"
ebs_block_device.2554893574.delete_on_termination: "true" => "true" (forces new resource)
ebs_block_device.2554893574.device_name:           "/dev/sdc" => "/dev/sdc" (forces new resource)
ebs_block_device.2554893574.encrypted:             "false" => "<computed>"
ebs_block_device.2554893574.iops:                  "150" => "<computed>"
ebs_block_device.2554893574.snapshot_id:           "" => "<computed>"
ebs_block_device.2554893574.volume_size:           "50" => "50" (forces new resource)
ebs_block_device.2554893574.volume_type:           "gp2" => "gp2" (forces new resource)
[Other three EBS blocks show the same messages.]

One explanation is that the AMI I'm using (Amazon ECS Optimized, ami-6df8fe7a), defines two block devices. The output from aws ec2 describe-images --image-id ami-6df8fe7a:

{
    "Images": [
        {
            "VirtualizationType": "hvm", 
            "Name": "amzn-ami-2016.09.c-amazon-ecs-optimized", 
            "Hypervisor": "xen", 
            "ImageOwnerAlias": "amazon", 
            "EnaSupport": true, 
            "SriovNetSupport": "simple", 
            "ImageId": "ami-6df8fe7a", 
            "State": "available", 
            "BlockDeviceMappings": [
                {
                    "DeviceName": "/dev/xvda", 
                    "Ebs": {
                        "DeleteOnTermination": true, 
                        "SnapshotId": "snap-441eb8ad", 
                        "VolumeSize": 8, 
                        "VolumeType": "gp2", 
                        "Encrypted": false
                    }
                }, 
                {
                    "DeviceName": "/dev/xvdcz", 
                    "Ebs": {
                        "DeleteOnTermination": true, 
                        "Encrypted": false, 
                        "VolumeSize": 22, 
                        "VolumeType": "gp2"
                    }
                }
            ], 
            "Architecture": "x86_64", 
            "ImageLocation": "amazon/amzn-ami-2016.09.c-amazon-ecs-optimized", 
            "RootDeviceType": "ebs", 
            "OwnerId": "591542846629", 
            "RootDeviceName": "/dev/xvda", 
            "CreationDate": "2016-12-07T18:14:59.000Z", 
            "Public": true, 
            "ImageType": "machine", 
            "Description": "Amazon Linux AMI 2016.09.c x86_64 ECS HVM GP2"
        }
    ]
}

The line ebs_block_device.#: "5" => "4" makes me think that one of the AMI-defined blocks is designed as the root block and that the other is considered a standard EBS block. But since that second block is not tracked in the configuration, every plan or apply sees only four managed blocks in my configuration and find five blocks exist in AWS. So it resets all the blocks to the configuration definition.

The workaround seems to be to define the EBS devices of the AMI (beyond the first) in my configuration, by adding this:

ebs_block_device {
  device_name = "/dev/xvdcz"
  volume_size = 22
  volume_type = "gp2"
}

That fixes the issue, but I have to remember to update it if the AMI change. Less than ideal, unfortunately, since I grabbing the latest.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment