|arkiver e183dd0578||преди 2 години|
|megawarc @ 1b3f680951||преди 2 години|
|.gitignore||преди 2 години|
|.gitmodules||преди 8 години|
|Dockerfile||преди 2 години|
|README.md||преди 2 години|
|chunk-multiple||преди 2 години|
|chunker||преди 2 години|
|config.example.sh||преди 2 години|
|docker-boot.sh||преди 2 години|
|du-all||преди 2 години|
|offload-multiple||преди 2 години|
|offload-one||преди 2 години|
|pack-multiple||преди 2 години|
|pack-one||преди 2 години|
|upload-multiple||преди 2 години|
|upload-one||преди 2 години|
Some scripts to bundle Archive Team uploads and upload them to Archive.org. Use at your own risk; the scripts will need per-project adjustment.
These scripts make batches of uploaded warc.gz files, combine them into megawarcs and upload them to their permanent home on Archive.org.
Three processes work together to make this happen:
The chunker moves uploaded warc.gz files from the upload directory to a batch directory. When this directory has grown to 50GB, the chunker begins a new directory and moves the completed directory to the packing queue.
There can only be one chunker per upload directory. Chunking doesn’t take long, if the files are not moving to a different filesystem.
The packer monitors the packing queue. When the chunker brings a new directory, the packer removes the directory from the queue and starts converting it into a megawarc (using the megawarc utility). When that is done, the packer moves the megawarc to the upload queue and removes the original warc files.
If necessary, multiple packers can work the same queue. Packing involves lots of gzipping and takes some time.
The uploader monitors the upload queue. When the packer brings a new megawarc, the uploader removes the megawarc from the queue and uploads it to Archive.org. If the upload is successful, the uploader removes the megawarc.
If necessary, multiple uploaders can work the same queue.
The offloader monitors the upload queue. Instead of uploading to Archive.org, the megawarc will be sent to another host via rsync. This is useful when Archive.org has issues.
This can be used at the same time as the uploader without issues.
From the chunker to the uploader, the chunks move through the system as timestamped directories, e.g., 20130401213900.) This timestamp will also be used in the name of the uploaded item on Archive.org. The queues are directories. Processes ‘claim’ a chunk by moving it from the queue directory to their working directory. This assumes that
mv is an atomic operation.
For efficiency and to maintain the atomicity of
mv, the filesystem of the directories is very important:
Filesystems 1 and 2 do not have to be the same.
Create a configuration file called
config.sh and place it in the directory where you start the scripts. See the
config.example.sh for more details.
Run the scripts in
tmux or something similar.
touch RUN before you start the scripts. Use
rm RUN to stop gracefully.
./chunk-multiple(run exactly one)
./pack-multiple(you may run more than one)
./upload-multiple(you may run more than one)
./offload-multiple(you may run more than one, can work in tandem with
du -hsin all queues
The packing script will use all your I/O capacity. Consider using
ionice to run in at a lower priority, so it doesn’t hinder your incoming Rsync or outgoing curl uploads.
ionice -c 2 -n 6 nice -n 19 ./pack-multiple
The scripts are designed not to lose data. If a script dies, you can look in its working directory for in-progress items and move them back to the queue.
These scripts use Bash and Curl.
You should clone https://github.com/ArchiveTeam/megawarc to the
megawarc/ subdirectory of these scripts. The megawarc utility requires Python and Gzip.