📮 Circadian SCP Upload

Resumable, interruptible, SCP upload client for any files or directories generated day by day.

Use Case

You have a local directory that generates data every day on your local machine. The directory looks like this:

📁 data-directory-1
├── 📁 20190101
│   ├── 📄 file1.txt
│   ├── 📄 file2.txt
│   └── 📄 file3.txt
└── 📁 20190102
    ├── 📄 file1.txt
    ├── 📄 file2.txt
    └── 📄 file3.txt

... or like this:

📁 data-directory-2
├── 📄 20190101.txt
├── 📄 20190102-a.txt
├── 📄 20190102-b.txt
└── 📄 20190103.txt

... and you want to upload that data to a server, but only after the day of creation. Additionally, you want to mark the directories as "in-progress" on the remote server, so that the next processing steps will not touch unfinished days of data while uploading.

This tool uses SCP via the Python library paramiko to do that. It will write files named .do-not-touch in the local and remote directory during the upload process and deletes them afterwards.

Below, you can find a code snippet that defines a specific directory/file naming scheme (for example %Y%m%d-(\.txt|\.csv)). The tool uses this information to tell, when a specific file or directory has been generated. It will only upload files when at least one hour of the following day has passed.

Can't I use rsync or a similiar CLI tool for that?

Yes, of course. However, the actual copying logic of individual files or directories is just 130 lines of code of this repository. The rest of this library is dedicated to it being a plug-and-play solution for your codebase: logging, regex filters, interruptibility, in-progress markers, etc..

You should be able to pip install/poetry add/... and call a well documented upload client class instead of having to manually connect your codebase to rsync and do all of this pattern logic yourself.

How did you make sure, that the upload works correctly?

First of all, the whole codebase has type hints and is strict checked with Mypy - even the snippet in the usage section below is tye checked with Mypy.

Secondly, the date patterning is tested extensively and the upload process of the files and directories is tested with an actual remote server by generating a bunch of sample files and directories and uploading them to that server. You can check out the output of the test runs in the GitHub Actions of this repository - in the "Run pytests" step.

Thirdly, after the upload, the checksum of the local and the remote directories/files is compared to make sure that the upload was successful. Only if those checksums match, the local files are deleted.

Usage

Install into any Python ^3.10 project:

pip install circadian_scp_upload
# or
poetry add circadian_scp_upload

Configure and use the upload client:

import circadian_scp_upload

# Use the callbacks to customize the upload process
# and integrate it into your own application. All callbacks
# are optional and the callback object does not need to be
# passed to the upload client. The lambda functions below
# are the default values.

upload_client_callbacks = circadian_scp_upload.UploadClientCallbacks(
    # which directories to consider in the upload process; only supports
    # %Y/%y/%m/%d - does not support parentheses in the string
    dated_directory_regex=r"^" + "%Y%m%d" + r"$",

    # which files to consider in the upload process; only supports
    # %Y/%y/%m/%d - does not support parentheses in the string
    dated_file_regex=r"^.*" + "%Y%m%d" + r".*$",

    # use your own logger instead of print statements
    log_info=lambda message: print(f"INFO - {message}"),
    log_error=lambda message: print(f"ERROR - {message}"),

    # callback that is called periodically during the upload
    # process to check if the upload should be aborted
    should_abort_upload=lambda: False,
)

# teardown happens automatically when leaving the "with"-block
with circadian_scp_upload.RemoteConnection(
    "1.2.3.4", "someusername", "somepassword"
) as remote_connection:

    # upload a directory full of directories "YYYYMMDD/"
    circadian_scp_upload.DailyTransferClient(
        remote_connection=remote_connection,
        src_path="/path/to/local/data-directory-1",
        dst_path="/path/to/remote/data-directory-1",
        remove_files_after_upload=True,
        variant="directories",
        callbacks=upload_client_callbacks,
    ).run()

    # upload a directory full of files "YYYYMMDD.txt"
    circadian_scp_upload.DailyTransferClient(
        remote_connection=remote_connection,
        src_path="/path/to/local/data-directory-2",
        dst_path="/path/to/remote/data-directory-2",
        remove_files_after_upload=True,
        variant="files",
        callbacks=upload_client_callbacks,
    ).run()

You will get an informational output whereever you direct the log output to - the progress is only logged at steps of 10%:

INFO - 2015-01-16: starting
INFO - 2015-01-16: 5 files missing in dst
INFO - 2015-01-16: created remote directory at /tmp/circadian_scp_upload_test_1692833117/20150116
INFO - 2015-01-16:   0.00 % (1/5) uploaded
INFO - 2015-01-16:  20.00 % (2/5) uploaded
INFO - 2015-01-16:  40.00 % (3/5) uploaded
INFO - 2015-01-16:  60.00 % (4/5) uploaded
INFO - 2015-01-16:  80.00 % (5/5) uploaded
INFO - 2015-01-16: 100.00 % (5/5) uploaded (finished)
INFO - 2015-01-16: checksums match
INFO - 2015-01-16: finished removing source
INFO - 2015-01-16: done (successful)
INFO - 2010-03-25: starting
INFO - 2010-03-25: 5 files missing in dst
INFO - 2010-03-25: created remote directory at /tmp/circadian_scp_upload_test_1692833117/20100325
INFO - 2010-03-25:   0.00 % (1/5) uploaded
INFO - 2010-03-25:  20.00 % (2/5) uploaded
INFO - 2010-03-25:  40.00 % (3/5) uploaded
INFO - 2010-03-25:  60.00 % (4/5) uploaded
INFO - 2010-03-25:  80.00 % (5/5) uploaded
INFO - 2010-03-25: 100.00 % (5/5) uploaded (finished)
INFO - 2010-03-25: checksums match
INFO - 2010-03-25: finished removing source
INFO - 2010-03-25: done (successful)