glad pollen
Hello,
I wanted to clarify how cache is calculated / invalidated for directories from the host.
I've talked about this a bit in #python-sdk-dev🐍, but I have a use case where I'm importing objects into terraform. In practice, this means chaining commands, but each individual step needs to be allowed to fail, so the execs are happening in a loop with the state file being exported out on success, and imported in by the next step.
I'm conditionally (because the file may not exist) mounting the state using this snippet:
async def terraform_state(client: dagger.Client):
logger.info('Checking entries')
tf_dir = client.host().directory("terraform")
entries = await tf_dir.entries()
def add_tf_state(image: dagger.Container) -> dagger.Container:
if 'terraform.tfstate' in entries:
logger.info('Found terraform state! ')
image = image.with_file('/terraform/terraform.tfstate',
tf_dir.file("terraform.tfstate"))
return image
return add_tf_state
which I the use in a loop as:
for tf_identifier, resource_id in generate_lambda_imports():
try:
result = (
tf_image
.with_(await terraform_state(client))
.with_exec(
['-chdir=/terraform', 'import',
tf_identifier, resource_id],
)
.file("/terraform/terraform.tfstate")
)
success = await result.export("terraform/terraform.tfstate")
if result:
logger.info(
f"[green]Imported {tf_identifier} {resource_id}![/]"
)
except dagger.ExecError as err:
result = err.stderr
if 'already managed' in result:
logger.info(
f"[yellow]{tf_identifier} already imported!"
)
else:
logger.error(
f"[bold red]Caught error importing {tf_identifier}: " +
f"[/]\n {result}"
)
success = False
But what I'm seeing is, despite exporting a file out to the directory terraform and adding a new entry, the call to client.host().directory('terraform') is cached, so each step in the loop doesn't get the state from the previous one.
I could of course replace theclient.host().directory logic with something Python native, I'm just wondering why my intuition is wrong here 🙁
limpid prism
👋 client.host().directory is only performed once for the session and then additional calls will give you the same reference you originally created. I think what you want instead of exporting things to the host is to use a cacheVolume between the different loops so the tfstate is shared between of all them.
glad pollen
Ahaa, a cache volume is a good idea! I'm already using one for the terraform providers and modules, to speedup subsequent runs
Great thanks, that's very clear
limpid prism
, we're here if you have further questions.
happy hacking 🧑💻
glad pollen
Sorry, one further question; can I mount / copy files into a cache volume to 'hydrate it'?
limpid prism
Sorry, one further question; can I mount / copy files into a cache volume to 'hy...
before the pipeline starts you mean? Yes, you can just create a hydration pipeline that just mounts that volume and copies whatever it needs there
glad pollen
Right, I see; within a container, I can use with_exec to movefiles in / out of the cache, but I can't add to the cache using with_file or pull things out using .file("/cache_mount_point/my_cool_file") for example.