#terraform-tool

the funny part is that it used to work last night

then today it stopped working all of a sudden

im not sure what i’m doing wrong

crolek: because you've changed the state of what terraform thinks that instance definition should be.

@pll, terraform is NOT considering it a change, thats the problem.

Oh, did I misread your question?

Oh, I did.

so if it's inline = [ "touch somefile"] and we change it to inline = [ "echo something"]

My apologies.

it considers that "not a change"

which i'd expect it to destroy the instance and make a new one (if it was an aws_instance)

(no problem pll :) )

Hmm, perhaps because a remote-exec is essentially packaged up as a shell-script and lobbed over to the other side.

oh, and its the "same" shell script under the covers?

Since the shell-script is still there, the terraform definition hasn't changed, only the contents of the script.

Yeah, I think so.

any idea how to force that update?

(i mean it makes sense what youre saying.)

hmm. I think you can taint the object, then re-run the apply.

That should force it.

dang, that's not what i was hoping to hear :/

Why ?

im thinking of just local execing now

Oh, a taint will require a destroy rebuild, right?

this is ridiculous

the azurerm ssh agent code is buggy :|

yeah it will, but i'd have to manually taint a lot of resources to potentially update one inline script

hmm.

but i think i can figure out a better solution, i just now have a better idea of "why" its not considering it a chagne

allfro: haven't used azure at all. I'm surprised the ssh code would be any different than the aws ssh code.

it’s not as far as i can tell

looking at the code it looks like it goes back down to a common file provisioner

so really the file provisioner code appears to be buggy

or the ssh agent code from the thirdparty is

allfro: interesting.

but so far i have had no luck today doing remote-exec or file provisioner

i keep getting ssh handshake errors

yet i see no outbound ssh being initiated

thank you again pll for helping answer my question(s)! :D

crolek: Perhaps if you comment out the entire remote-exec, run apply, uncomment, run apply again ?

agent or agentless ssh fails

:(

crolek: anytime. I hope you figure it out.

allfro: can you set ssh -v to see where/why?

how do i do that in terraform

it’s terraform that is calling ssh somehow

good question...

im doing TF_LOG=DEBUG

but i get nothing on what ssh is doing

hmm, maybe if you set it in your .ssh/config ?

all i see is the end result which is that the handshake fails

let me see if that does anything

in your .ssh/config you can set: LogLevel VERBOSE|DEBUG1|DEBUG2|DEBUG3

pll i did that

and im experimenting

but i dont see the ssh verbosity increasing

hmmm. That's strange.

out of curiosity, does the verbosity increase using a normal ssh session ?

?

not sure what you mean pll

if you just run ssh from the command line with LogLevel set to VERBOSE, do you see a verbose session?

n

no

i dont see a verbose session unfortunately :(

bummer.

I've gotta run, good luck figuring this out.

nvm

VERBOSE is not what i thought it would be

has to be DEBUG

oh, that makes sense.

anybody else having a lot of pain with the remote exec and file provisioner for linux hosts?

uhm

why is terraform trying to connect to localhost?

when remote provisioning

?

hey everybody

Does anyone have an example of creating ebs volumes and attaching them to servers spun up from an autoscaling group?

SneakyPhil: https://www.terraform.io/docs/providers/aws/

have you taken a look at the different resource provisioners?

in the docs

Hi, I'm a new Terraform user and I'm having trouble solving an issue that none of my colleagues can answer. I am trying to move a service over to a new server and I need to create new load balancers, but I want to use an existing DNS record that wasn't created through Terraform. What is happening is that when I create the new load balancer, its recreating the DNS record and there is about 5-10 minutes downtime in which I cannot r

this migration with no downtime? Can I somehow add the existing DNS entry into Terraform even though it was created through the AWS Console?

chester, look into the import feature