Skip to content

git001/caddyv2-upload

Repository files navigation

caddyv2-upload

This repo holds a simple caddyserver v2 upload handler

Warning
You have to secure the upload URL as the nature of this handler/plugin is to save uploaded artifacts on the disc.

In this handler are the following modules in use.

  • templates

  • file_server

  • github.com/caddyserver/jsonc-adapter

  • upload :-)

Configuration

Important
The relative paths are relative to the current web root directory http.vars.root!

Environment Variables

Table 1. Parameters
Name Description

APPPORT

The Port on which Caddyserver should listen. The format depends which Configuration file format you use. See the official Caddyserver documentation for the format Network addresses

SKIP_LOG

With this variable can you control if the health check will be logged or not. This feature was implemented as part of the vars handler.

Parameters

Table 2. Parameters
Name Description

dest_dir

The directory in which the files will be uploaded

root_dir

The root directory for this module. When not set will the root directory be evaluated at runtime from this variable http.vars.root.

file_field_name

The field name for the multi-part form upload

fixed_file_name

When set, all uploaded files will be renamed to the provided file name.

response_template

The response template after a upload was successfully

max_form_buffer

The maximum buffer size for ParseMultipartForm. It accepts all size values supported by go-humanize. Here can also be used a Placeholder
JSON: "{env.MAXBUFFSIZE}"
Caddyfile: {$MAXBUFFSIZE}

The default size is 1G.

max_form_buffer_int

The maximum buffer size for ParseMultipartForm.
The default size is 1G.

max_filesize

is the maximum size in bytes allowed for the upload. It accepts all size values supported by go-humanize. Reads of more bytes will return an error with HTTP status 413. Here can also be used a Placeholder
JSON: "{env.MAXFILESIZE}"
Caddyfile: {$MAXFILESIZE}

The default size is 1G.

max_filesize_int

is the maximum size in bytes allowed for the upload. Reads of more bytes will return an error with HTTP status 413.
The default size is 1G.

notify_url

After a successful upload will this URL be called. The only supported schema is https.

notify_method

The default method is GET. If you need to make a POST request please open a feature issue as for now is a request body handling not implemented

insecure

This boolean flag configure the InsecureSkipVerify in the tls-config . Default is false which implies that a valid CA must be used

capath

This is the Parameter where you can define the CA filename for the notify_url.

create_uuid_dir

If set to true, each file will get a unique directory with a UUID as its name.

dest_dir_field_name

If set, the dest_dir will be set from the specified form value.

pass_thru

If set to true, enables pass-thru mode, which continues to the next HTTP handler in the route instead of responding directly.

Caddy Variables

The following Variables will be added to the request flow

Table 3. variables
Name Description

http.upload.filename

The uploaded filename

http.upload.filesize

The uploaded filesize

http.upload.directory

The directory where the file is uploaded

http.upload.uuiddir

The name of the random directory created if create_uuid_dir is enabled

JSON

Because I prefer the JSON Config will I write here the configuration snipplet in JSON Syntax.

	"handle": [
		{
			"MyTlsSetting": {},
			"dest_dir": "upload", (1)
			"handler": "upload",
			"max_filesize": "5GB", (2)
			"max_form_buffer_int": 1000000, (3)
			"response_template": "upload-resp.txt" (4)
		}
	]
  1. Destination Directory on the Server site

  2. Maximal possible upload size

  3. Maximal buffer for uploading

  4. the response template which will be used for response after upload

A full working example is in docker-files/opt/webroot/config/Caddyfile-upload.json

Caddyfile

Here a example Caddyfile which expects that the environment variable APPPORT is set.

{
	order upload before file_server
	log {
		level DEBUG
	}
}

{$APPPORT} {
	root .

	file_server browse
	templates

	@mypost method POST
	upload @mypost {
		dest_dir tmp-upl
		max_form_buffer 1G
		max_filesize 4MB
		response_template templates/upload-resp-template.txt
	}

	log {
		output file access.log
	}
}

build

local

xcaddy build --with github.com/kirsch33/realip \
  --with github.com/git001/caddyv2-upload

docker

buildah bud --tag caddyv2-upload .
# or
docker build --tag caddyv2-upload .

run

cli

APPPORT=:2011 ./caddy run \
  -config Caddyfile-upload.json

docker

You can get this image from docker hub

The default listen port must be defined with this variable

APPPORT=:2011

podman run --rm --network host --name caddy-test \
  --env APPPORT=:8888 -it \
  docker.io/me2digital/caddyv2-upload:latest
# or
docker run --name caddy-test --rm \
  docker.io/me2digital/caddyv2-upload:latest

OpenShift / kubernetes

You can use the examples in the directory openshift to deploy the caddy uploader into your OpenShift or kubernetes Cluster.

oc new-project caddyupload
oc -n caddyupload apply -f openshift/

To adopt the used Caddyfile then can you easily do this with a configmap. The commands below show the steps.

# add configure change trigger
oc -n caddyupload \
  set triggers \
  deployment.apps/caddy-upload \
  --from-config

# create configmap
oc -n caddyupload \
  create configmap \
  caddyfile-json \
  --from-file=Caddyfile-upload.json=<PATH_TO_LOCAL>/Caddyfile-upload.json

# add configmap to caddy deployment
oc -n caddyupload \
  set volumes \
  deployment.apps/caddy-upload \
  --add --configmap-name=caddyfile-json \
  --mount-path=/opt/webroot/config/Caddyfile-upload.json \
  --name=caddyfile-json \
  --sub-path=Caddyfile-upload.json \
  --type=configmap

In case the first shot of the configuration file does not work can you use the following commands to recreate the configuration file.

oc -n caddyupload-nis delete configmap caddyfile-json \
&& oc -n caddyupload-nis create configmap caddyfile-json --from-file=Caddyfile-upload.json=<PATH_TO_LOCAL>/Caddyfile-upload.json

# output of the commands
configmap "caddyfile-json" deleted
configmap/caddyfile-json created

Health probe

There is a builtin health handler with the path /health which just returns 200 and Okay. The log output can be controled via the environment variable SKIP_LOG.

example cli

When you run the Image with port 8888 can you use curl or any other tool to post (upload) files

It’s not necessary to use -X POST as written in this Blog post UNNECESSARY USE OF CURL -X

Here a example call with curl

curl -v --form [email protected] http://localhost:8888/templates/upload-template.html
*   Trying 127.0.0.1:8888...
* TCP_NODELAY set
* Connected to localhost (127.0.0.1) port 8888 (#0)
> POST /templates/upload-template.html HTTP/1.1
> Host: localhost:8888
> User-Agent: curl/7.68.0
> Accept: */*
> Content-Length: 2492
> Content-Type: multipart/form-data; boundary=------------------------58b770bc61c0e691
> Expect: 100-continue
>
* Mark bundle as not supporting multiuse
< HTTP/1.1 100 Continue
* We are completely uploaded and fine
* Mark bundle as not supporting multiuse
< HTTP/1.1 200 OK
< Accept-Ranges: bytes
< Content-Length: 299
< Etag: "rbb1gx8b"
< Last-Modified: Tue, 03 May 2022 11:34:09 GMT
< Server: Caddy
< Date: Thu, 19 May 2022 21:45:07 GMT
<

http.request.uri.path: {{placeholder "http.request.uri.path"}}

http.request.uuid {{placeholder "http.request.uuid" }}
http.request.host {{placeholder "http.request.host" }}

http.upload.filename: {{placeholder "http.upload.filename"}}
http.upload.filesize: {{placeholder "http.upload.filesize"}}

Background informations

The max_form_buffer paramater will be directly passed to readForm function and is used to check if the uploaded file should be saved temporarly on disk or keep it in the memory. This have dicret impact into the performance and disk usage of that module. Keep in mind when this paramter is low and the upload is a big file then will be there a lot of disk io.

INFO: The observation from https://github.com/etherwvlf in issue Memory issues on large uploads was that the initial memory usage is 7-8 times higher then the configured max_form_buffer size.