Froster is a user-friendly archiving tool for teams that move data between high-cost POSIX file systems and low-cost S3-like object storage systems. It currently supports these S3 providers: AWS, GCS, Wasabi, IDrive, Ceph, and Minio. Froster can efficiently crawl your Posix file system metadata, recommend folders for archiving, generate checksums, and upload your selections to Glacier or other S3-like storage. It can retrieve data back from the archive using a single command. Additionally, Froster can mount S3/Glacier storage inside your on-premise file system and also restore it to an AWS EC2 instance.

Installation pre-requisite: packages

On Debian / Ubuntu

sudo apt update
sudo apt install -y curl python3 python3-pip python3-venv python3-dev gcc lib32gcc-s1 unzip fuse3

On Fedora / CentOS / RHEL

sudo dnf update -y
sudo dnf groupinstall -y "Development Tools"
sudo dnf install -y curl python3 python3-pip python3-devel unzip fuse3

On HPC machine

Please contact your administrator to install these packages:

curl python3 python3-pip python3-venv python3.xx-devel gcc lib32gcc-s1 unzip fuse3

Installation

To install Froster, execute the following command in your terminal:

curl -s https://raw.githubusercontent.com/dirkpetersen/froster/main/install.sh?$(date +%s) | bash

To check that everything is installed correctly, run froster --info or froster --version commands. Check available command on the help section froster --help

Update

To update Froster, execute the same command from the Installation section.

Froster prompts you once a week when there is a new update available. You can manually check if there is an update available by running the command:

froster update

Configuration

To run Froster, you need to configure it first. Froster will guide you through the configuration with questions you need to answer. In most cases, you can accept default options by pressing enter. To configure Froster, run the command:

froster config

If you have any doubts while configuring Froster, check the Configuration sections explanation

Import and Export Configuration

If several users of the same team wants to use Froster, they will have to configure the same profiles to work with same configuration, same buckets, same endpoints, etc. To ease this configuration accross teams a user that has already configured Froster can export its configuration. This will export all configuration except for sensitive information: username, email and credentials. To export a configuration run the command:

froster config --export /folder/path

To import a configuration, run the command:

froster config --import /path/to/exported/config

Please note that froster config command still needs to be executed, as username, email and credentials still need to be configured.

Basic usage of Froster

Credentials

The credentials commands checks that the current selected profile is correctly configured and has valid credentials.

froster credentials

Index

The index command crawls your file system, looking for hotspots (i.e., folders with files larger than 1GB). It uses a third-party tool called filesystem-reporting-tools, which uses parallel threads to crawl large filesystems efficiently.

This command is not strictly necessary for archiving, but it helps you to spot large folders that are potentially worth archiving. You can provide one or more folders separated by space:

froster index [folders...]

Check more options at froster index --help.

Archive

The archive command uploads folders from the local filesystem to the default profile's S3 bucket. If one or more folders are provided, they will be archived. If no folder is provided, froster will allow you to select hotspots of previously indexed folders.

froster archive [folders...]

Check more options at froster archive --help.

Note: To change default profile use froster --default-profile command. To use a different profile for the current command, you can use the --profile PROFILE flag. Check these options at froster --help

Delete

The delete command deletes previously archived folders. If one or more folders are provided, all of them will be deleted from the local filesystem ONLY after verifying that they have been successfully stored in the S3 server. If no folder is provided, froster will allow you to select a folder from previously successfully archived folders. It is worth noting that this commands checks if the folders have been successfully upload to the S3 bucket. Please do NOT delete folders manually.

froster delete [folders...]

Check more options at froster delete --help.

Restore

The restore command restores a previously deleted folder (i.e., deleted by using the froster delete command). This command will first trigger the folder retrieval if folders are stored in a DEEP_ARCHIVE storage class (See AWS Storage Class guide). Once retrieval finishes, executing this command again will download files from the S3 bucket to the local filesystem.

froster restore [folders...]

Check more options at froster restore --help.

Mount / Umount

The mount / umount command (u)mounts the remote S3 or Glacier storage in your local file system at the location of the original folder. If one or more folders are provided, they will be (u)mounted. If no folder is provided, froster will allow you to select folders of previously archived folders.

froster mount [folders...]

froster umount [folders...]

Check more options at froster mount --help.

Table of Contents

• Problem
  • Motivation
• Design
• Preparing Froster
  • Configuring
  • Changing defaults with aliases
  • Working with multiple users
    • AWS configuration for teams
  • Advanced configuration
• Using Froster
  • Standard usage
  • Large scale use on HPC
  • Picking old and large data
  • Special use cases
  • Recursive operations
  • Tarring small files
  • NIH Life Science metadata
  • Restore to cloud machine (EC2)
  • Using desktop tools to browse S3 Glacier
    • Cloudberry Explorer
    • Cyberduck
  • More detailed file system analysis
• Command line help
• FAQ and Troubleshooting
• Commercial Solutions
• Discontinuing Froster

Problem

We have yet to find an easy to use, open-source solution for large-scale archiving intended to free up disk space on a primary storage system. Researchers, who may have hundreds of terabytes or even petabytes of data, need to decide what to archive and where to archive it. Archiving processes can stretch on for days and are susceptible to failure. These processes must be able to resume automatically until completion, and need to be validated (for instance, through a checksum comparison with the source). Additionally, it's crucial to maintain some metadata regarding when the archiving occurred, where the data was transferred to, and the original location of the data to be able to restore it quickly.

Motivation

The main motivations behind the creation of Froster are:

• We were working with researchers to optimize their archiving processes. In doing so, we wanted to understand user workflows and identify the essential metadata to be captured. Thanks to Froster's compact codebase, it can be easily adapted to various needs. Metadata could potentially be captured through an additional web interface. However, a web system often requires IT support and an extra layer of authentication. It might be simpler to gather such information through a Textual TUI interface. A first demo of this option is here. The insights gained through working with Froster will likely guide the development or acquisition of more advanced tools in the future.

• HPC users often find AWS Glacier inaccessible or complex and have a preception that it is expensive because of Egress fees. This is despite evidence that most data for most research users has not been touched in years. It seems that it is very difficult to build an on premises archive solution with a lower TCO than AWS Glacier or similar offerings from other cloud providers such as IDrive E2

• Secure the archiving process via checksums verification. After every transfer, a checksum verifies the integrity of destination files, ensuring no data gets lost.

Design

1. First, we may need to crawl the file system that sometimes has billions of files to find data that is actually worth archiving. For this, Froster uses the well-known pwalk, a multi-threaded parallel file system crawler that creates a large CSV file of all discovered file metadata.

2. Our focus is on archiving folders rather than individual files, as data that resides in the same folder typically belongs together. To this end, we filter the pwalk's CSV file with DuckDB, and get a new list (CSV) of the largest folders sorted by size. (We call this a Hotspots file)

3. We pass this new list (CSV) to an interactive tool based on Textual, which displays a table of folders alongside their total size in GiB, their average file sizes in MiB (MiBAvg), as well as their age in days since last accessed (AccD) and modified (ModD). Users can scroll down and right using the mouse, and hitting enter will trigger the archiving of the selected folder.

4. All files < 1 MiB size are moved to an archive called Froster.smallfiles.tar prior to uploading. You can also avoid tarring

5. We pass the folder to Rclone to execute a copy job in the background.

6. Many researchers with very large datasets will have a Slurm cluster connected to their data storage system. This is ideal for long-running data copy jobs, and Slurm can automatically re-run copy jobs that failed. Researchers can also use the --no-slurm option to execute the job in the foreground. This is the default if Slurm is not installed on the Linux machine.

7. Prior to copying, we place files with checksums (e.g., .froster.md5sum) in the folders to be archived. This allows for an easy subsequent checksum comparison, providing the user with evidence that all data was correctly archived (through size and md5sum comparisons).

8. After files have been deleted from the source folder, we place a file named Where-did-the-files-go.txt in the source folder. This file describes where the data was archived, along with instructions on how to retrieve the data.

9. Optionally allow archive --recursive to enable archiving of entire folder trees. More details here.

10. Optionally link to NIH metadata with the archive --nih option to link to life sciences research projects More details here.

Preparing Froster

Configuring

The config is mostly automated. You can just run froster config command. You need to answer a few questions for which you can accept defaults in most cases.

When running froster config, you should confirm the default DEEP_ARCHIVE for AWS S3 storage class as this is currently the most cost-effective storage solution available. It takes only 48-72 hours to retrieve your data with the least expensive retrieval option ('Bulk'). See AWS Storage Class guide. However, you can choose other AWS S3 storage classes supported by the rclone copy tool.

Froster uses ~/.aws files to store credentials, regions, and endpoints in the AWS standard way.

Check the rclone S3 docs to learn about different providers and endpoints.

Froster has been tested with numerous S3-compatible storage systems such as GCS, Wasabi, IDrive, Ceph, and Minio. At this time, only AWS supports the DEEP_ARCHIVE storage class.

Configuration sections explanation

Down below, you will find an explanation of each configuration section:

• SET USER

  • Set your name. This information will be stored locally as metadata when archiving files.

• SET EMAIL

  • Set your email. This information will be stored locally as metadata when archiving files. Email is also used to notify about SLURM executions.

• SET SHARED

  • If set to yes, then enter the path to the shared folder. The shared folder will be used to store hotspots (output of froster index <folder> command) and the archiving database (file with all archived files). This is useful when Froster is used by more than one user. Current user hotspots will be copied to the new shared folder. The archiving database will ONLY be copied to the shared folder if no archiving database is already in the shared folder.
  • If set to no, hotspots and the archiving database will be stored in the local user filesystem.

• SET NIH

  • If set to yes, it allows you to create a link between the data you archive and the research project granted by a funding organization to increase the FAIR level of an archived dataset. The National Institutes of Health (NIH) has maintained a large database of all publicly funded life sciences projects in the US since the 1980s at NIH RePORTER.

• SET PROFILE

  • Profiles are the most important configuration of Froster. You will need to have at least one configured profile to be able to archive folders to your S3 provider. Profiles store the following information:
  • Select profile: Configure an already existing profile or create a new one.
  • Select S3 provider: Select the S3 provider for this profile. Currently supported S3 providers are AWS, GCS, Wasabi, IDrive, Ceph, and Minio. There is also a generic option called "Other," where you can configure a generic S3 provider.
  • Select credentials: Select the credentials for this profile or create new ones. These credentials will be used to authenticate to your S3 provider. They are stored in the ~/.aws/credentials file.
  • Select region: Select the region for this profile or create a new one. -- no region -- option will set the region to the value "default".
  • Select S3 bucket: Select the bucket for this profile or create a new one.
  • Enter directory inside S3 bucket: Enter the root folder where files will be archived for the selected S3 bucket.
  • Select the S3 storage class: Select the S3 storage class for this profile or create a new one. DEEP_ARCHIVE is currently the most cost-effective storage solution available. See AWS Storage Class guide

• SET SLURM

  • Configure several SLURM options that will be passed to the SLURM script upon execution.

AWS configuration for teams

You typically use AWS Identity and Access Management (IAM) to set up policies for S3. You can define permissions at a granular level, such as by specifying individual API actions or using resource-level permissions for specific S3 buckets and objects.

Here's a step-by-step guide for a simple permission setup for Data Stewards and Archive Users using the IAM Management Console:

1. Open the IAM console at https://console.aws.amazon.com/iam/ and login with an administrative user account.
2. Create User Groups: It's easier to manage permissions by grouping users. Create two groups: ReadWriteGroup and ReadOnlyGroup. Add the Data Stewards to ReadWriteGroup and the Archive Users to ReadOnlyGroup
3. Create one each IAM Policy (JSON Files) for ReadWriteGroup and ReadOnlyGroup. The s3:RestoreObject action allows users to restore data from Glacier.
4. Attach Policies to Groups: In the IAM console, select the ReadWriteGroup group, click on "Add permissions" and then "Attach policies". Create a new policy using the first JSON, and attach it to this group. Do the same for ReadOnlyGroup using the second JSON.

Policy ReadWriteGroup:

{
    "Version": "2012-10-17",
    "Statement": [
        {
            "Effect": "Allow",
            "Action": [
                "s3:PutObject",
                "s3:GetObject",
                "s3:ListBucket",
                "s3:DeleteObject",
                "s3:RestoreObject"
            ],
            "Resource": [
                "arn:aws:s3:::your-bucket-name",
                "arn:aws:s3:::your-bucket-name/*"
            ]
        }
    ]
}

Policy ReadOnlyGroup:

{
    "Version": "2012-10-17",
    "Statement": [
        {
            "Effect": "Allow",
            "Action": [
                "s3:GetObject",
                "s3:ListBucket",
                "s3:RestoreObject"
            ],
            "Resource": [
                "arn:aws:s3:::your-bucket-name",
                "arn:aws:s3:::your-bucket-name/*"
            ]
        }
    ]
}

Changing defaults with aliases

Froster has a number of global options that you may prefer but you don't like to type them all the time. We have --no-slurm (don't submit batch jobs even if Slurm is found), --cores (use a number of CPU cores that is different from the default 4) and --profile (use a different profile than default_profile for S3 related commands). You can create a command alias that uses these settings by default. For example, if you think that typing froster is too long and would like to rather type fro and use 8 cpu cores in the foreground with a custom profile named myceph you can simply enter this command in bash:

alias fro="froster --no-slurm --cores=8 --profile=myceph"

If you would like to keep this past the next login you can add this alias to the bottom of ~/.bashrc:

echo 'alias fro="froster --no-slurm --cores=8 --profile=myceph"' >> ~/.bashrc

Using Froster

Standard usage

In its most simple form you can use Froster to archive a specific folder, delete the source and restore the data interactively on a standalone Linux machine.

For example we have 2 files in a folder

vmachado@vmachado:~$ ls -la ~/Desktop/froster/csv

total 2008
-rw-rw-r--  1 vmachado vmachado 1126400 jul  1 17:19 9j6_3ne_medium_1.out
-rw-rw-r--  1 vmachado vmachado  921600 jul  1 17:19 9j6_rqu_small_1.out

Now let's archive this folder

vmachado@vmachado:~$ froster archive ~/Desktop/froster/csv

ARCHIVING /home/vmachado/Desktop/froster/csv

    Generating Froster.allfiles.csv and tar small files...
        ...done

    Generating checksums...
        ...done

    Uploading Froster.allfiles.csv file...
        ...done

    Uploading files...
        ...done

    Verifying checksums...
        ...done

ARCHIVING SUCCESSFULLY COMPLETED

    PROVIDER:           "AWS"
    PROFILE:            "profile aws"
    ENDPOINT:           "https://s3.eu-west-2.amazonaws.com"
    LOCAL SOURCE:       "/home/vmachado/Desktop/froster/csv"
    S3 DESTINATION:     ":s3:froster-vmachado-test/froster/home/vmachado/Desktop/froster/csv"

Froster creates md5 checksums for the files in the local folder, uploads the files and then compares the checksums in the archive with the local ones.

Now we delete the local data as we have evidence that the data is intact in the archive. We could execute froster delete ~/csv but, since the archiving process took a long time, we forgot which folder we needed to delete. We just run froster delete without the folder argument to see a list of archived folders and pick the one in which we want to delete files.

after the deletion is completed, we see that a manifest file was created that gives users some info.

vmachado@vmachado:~$ froster delete

DELETING /home/vmachado/Desktop/froster/csv...

    Verifying checksums...
        ...done

    Deleting files...
        ...done

DELETING SUCCESSFULLY COMPLETED

    LOCAL DELETED FOLDER:   /home/vmachado/Desktop/froster/csv
    AWS S3 DESTINATION:     :s3:froster-vmachado-test/froster/home/vmachado/Desktop/froster/csv

    Total files deleted:    2

    Manifest:               /home/vmachado/Desktop/froster/csv/Where-did-the-files-go.txt

The file has information that should allow other users to find out where the data went and who to contact to get it back. It is important to understand that before deletion, Froster will compare the checksums of each file that was archived with the file in the archive and will only delete local files that have an identical copy in the archive. The manifest file and the checksum comparison are good reasons why you should always let Froster handle the data deletion instead of deleting the source folder manually.

vmachado@vmachado:~$ cat /home/dp/csv/Where-did-the-files-go.txt

The files in this folder have been moved to an AWS S3 archive!

Archive location: :s3:froster-vmachado-test/froster/home/vmachado/Desktop/froster/csv
Archive aws profile: profile aws
Archiver user: vmachado
Archiver email: victor.machado@hpcnow.com
froster-archives.json: /home/vmachado/.local/share/froster/froster-archives.json
Archive tool: https://github.com/dirkpetersen/froster
Restore command: froster restore "/home/vmachado/Desktop/froster/csv"
Deletion date: 2024-07-01 17:23:19.917602

First 10 files deleted this time:
9j6_3ne_medium_1.out, Froster.smallfiles.tar

Please see more metadata in Froster.allfiles.csv file

You can use "visidata" or "vd" tool to help you visualize Froster.allfiles.csv file

After a folder has been deleted and you wish to view the file names in the archive, simply execute froster mount. Then, select the folder you want to access. Froster will then mount the bucket location to the folder that was deleted. This allows you to see file and folder names, along with their modification dates. Additionally, you can view the Froster.allfiles.csv file using any editor or the visidata / vd tool. This particular file contains metadata for all files in the folders, including those archived in a .tar format. If you haven't used the DEEP_ARCHIVE or GLACIER S3 archive tiers, you can also copy individual files or access them through your existing pipelines. Remember to use froster umount to unmount the folder when done.

vd /home/vmachado/Desktop/froster/csv
froster mount
froster umount

The visidata tool can display but also manipulate csv files:

After a while you may want to restore the data. Again, you forgot the actual folder location and invoke froster restore without the folder argument to see the same dialog with a list of achived folders. Select a folder and hit "Enter" to restore immediatelty.

vmachado@vmachado:~$ froster restore

RESTORING /home/vmachado/Desktop/froster/csv

    Triggered Glacier retrievals: 2
    Currently retrieving from Glacier: 0
    Retrieved from Glacier: 0
    Not in Glacier: 1

    Glacier retrievals pending. Depending on the storage class and restore mode run this command again in:
        Expedited mode: ~ 5 minuts

        Standard mode: ~ 12 hours

        Bulk mode: ~ 48 hours

NOTE: You can check more accurate times in the AWS S3 console

If you use the DEEP_ARCHIVE (default) or GLACIER AWS S3 storage classes, the first execution of froster restore will initiate a Glacier retrieval. This retrieval will copy the data in the background from the Glacier archive to the S3 One Zone-IA storage class which costs about $10 per TiB/month and keep it there for 30 days by default (you can change this to something like 7 days with froster restore --days 7). Wait for 12 hours and run the froster restore command again. If all data has been retrieved (which means both "Triggered Glacier retrievals" and "Currently retrieving from Glacier" show 0) the restore to the original folder location will proceed. As an alternative to running froster restore a second time, you can use froster mount which allows you to access individual files once they have been retrieved from Glacier.

dp@grammy:~$ froster restore

RESTORING /home/vmachado/Desktop/froster/csv

    Triggered Glacier retrievals: 0
    Currently retrieving from Glacier: 0
    Retrieved from Glacier: 2
    Not in Glacier: 1

    Downloading files...
        ...done

    Generating checksums...
        ...done

    Verifying checksums...
        ...done

    Untarring Froster.smallfiles.tar... 
        ...done

RESTORATION OF /home/vmachado/Desktop/froster/csv COMPLETED SUCCESSFULLY

Note that if you restore from AWS S3 to on-premises, you may be subject to AWS data egress charges (up to $90 per TiB downloaded, or $20 per TiB with AWS DirectConnect), unless your organization has negotiated an AWS data egress waiver.

Large scale use on HPC

If you have hundreds of terabytes or even petabtyes of data in billions of files you may not be able to easily locate data that is worth archiving. Among hundreds of thousands of folders there are typically only a few hundred that make up most of your storage consumption. We call these folders 'Hotspots' and to find them you use the froster index command and pass the root directory of your lab data.

[user@login ~]$ froster index /home/vmachado/data_2

SLURM JOB

  ID: 2251
  Type: index
  Check status: "squeue -j 2251"
  Check output: "tail -n 100 -f /home/.local/shared/froster/slurm/froster-index@+home+vmachado+data_2-2251.out"
  Cancel the job: "scancel 2251"

we can see the status of the job and then use the suggested tail command to see how the indexing is progressing, we can see that the job finished successfully

[user@login ~]$ tail -n 100 -f /home/.local/shared/froster/slurm/froster-index@+home+vmachado+data_2-2251.out

INDEXING /home/vmachado/data_2

Hotspots file: /home/vmachado/shared_dir/hotspots/+home+vmachado+data_2.csv
    with 0 hotspots >= 1 GiB
    with a total disk use of 0.0 TiB

Total folders processed: 3

INDEXING SUCCESSFULLY COMPLETED

now we run the froster archive command without entering folder names on the command line. We just created a csv folder for testing that contains 3 csv files with a total of about 1.1 GiB data. We pick this csv folder. It shows that it has been accessed 0 days ago (column AccD) because we just created it.

[user@login ~]$ froster archive /home/vmachado/data_2

SLURM JOB

  ID: 2252
  Type: archive
  Check status: "squeue -j 2252"
  Check output: "tail -n 100 -f /home/.local/shared/froster/slurm/froster-archive@+home+vmachado+data_2-2252.out"
  Cancel the job: "scancel 2252"

Once the archive job has completed you should receive an email from your Slurm system.

Picking old and large data

Now let's take it to the next level. Small datasets are not really worth archiving as the cost of storing them is low .... and if we started archiving data that was just used last week, some users would get really angry with us. This means our task is pretty clear: First we are looking for datasets that are large and old. In the example below we want to identify datasets that have not been touched in 3 years (1095 days) and that are larger than 1 TiB (1024 GiB without subfolders). We are executing froster archive --older 1095 --larger 1024. As you can see, this produces an output command that you can copy and paste directly into the shell of your HPC login node to submit an archiving batch job. In this case you will save more than 50 TiB of disk space with a few seconds of labor even though the archiving job may take days to run.

froster archive --older 1095 --larger 1024

Processing hotspots file /home/users/dp/.froster/config/hotspots/@XxxxermanLab.csv!

Run this command to archive all selected folders in batch mode:

froster archive \
"/home/groups/XxxxermanLab/xxxriti/tabulation-virus2" \
"/home/groups/XxxxermanLab/xxxsojd/KIR6.2_collaboration" \
"/home/groups/XxxxermanLab/xxxtofia/ER-alpha/WildType_Estradiol_Pose1_BrokenH12_WExplore" \
"/home/groups/XxxxermanLab/xxxriti/estrogen-receptor" \
"/home/groups/XxxxermanLab/xxxsojd/covid_we_data" \
"/home/groups/XxxxermanLab/xxxsojd/covid_data/we_data" \
"/home/groups/XxxxermanLab/xxxtossh/lc8-allostery-study/gromacs/7D35/7D35-comparison/singly-bound-structure/md_5us" \
"/home/groups/XxxxermanLab/xxxtossh/lc8-allostery-study/gromacs/7D35/7D35-comparison/doubly-bound-structure/md_5us" \
"/home/groups/XxxxermanLab/xxxtossh/lc8-allostery-study/gromacs/7D35/7D35-comparison/apo-structure/md_5us" \
"/home/groups/XxxxermanLab/xxxperma/cell/live_cell/mcf10a/batch_17nov20/wctm/17nov20" \
"/home/groups/XxxxermanLab/xxxperma/cell/live_cell/mcf10a/ligcomb_17jun21"

Total space to archive: 54,142 GiB

Note: you can also use --newer xxx --larger yyy to identify files that have only been added recently

Special use cases

Recursive operations

By default, Froster ignores sub-folders, as only files that reside directly in a chosen folder will be archived. There are two reasons for this: First, if your goal is cost reduction rather than data management, most folders are small and not worth archiving. We want to focus on the few large folders (Hotspots) that make a difference. Secondly, many storage administrators are uncomfortable with users moving hundreds of terabytes. If an entire folder tree can be archived with a single command, it may create bottlenecks in the network or storage systems. However, you can use the archive --recursive option with the archive command to include all sub-folders

Tarring small files

Folders with large amounts of small files have always been problematic to manage as they can degrade the metadata performance of a storage system and copying many of them is very slow. Since Glacier objects have an overhead of 40K per uploaded file, storage consumption can get quickly out of hand if not properly managed as a folder with 1 million tiny files would require an addional 40GB storage capacity.

Froster addresses this by moving all small files < 1 MiB to to a Tar archive called Froster.smallfiles.tar in each folder. Froster.smallfiles.tar is created prior to uploading. When restoring data Froster.smallfiles.tar is automatically extracted and then removed. You do not need to know anything about Froster.smallfiles.tar as you will only see Froster.smallfiles.tar in the S3 like objectstore after it has been deleted from the local file system. If you browse the objectstore with a tool such as Cyberduck you will see a file Froster.allfiles.csv along with Froster.smallfiles.tar. The CSV file contains a list of all filenames and metadata including the ones that were tarred to provide addional transparency.

To avoid tarring you can set max_small_file_size_kib to 0 using this command. The default is 1024 (KiB) or you can use the archive --notar option.

echo 0 > ~/.froster/config/general/max_small_file_size_kib

NIH Life Sciences metadata

In many cases we would like to create a link between the data we archive and the research project granted by a funding organization to increase the FAIR level of an archived dataset. The National Institutes of Health (NIH) maintains a large database of all publicly funded life sciences projects in the US since the 1980s at NIH RePORTER. Froster uses the RePORTER API to allow you to search for grants and link them to your datasets.

To invoke the search interface configure it as default on the froster config SET_NIH section or use the --nih option with archive subcommand

froster archive --nih [folders...]

You can search multiple times. Once you found the grant you are looking for hit TAB and use the allow keys to select the grant and confirm with enter

Using desktop tools to browse S3 Glacier

Cloudberry Explorer

Cloudberry Explorer is a GUI tool to browse, upload and transfer data from AWS S3/Glacier.

• Download Cloudberry Explorer for Windows or Mac
• Launch and select Amazon S3

• Enter your credentials and pick your preferred region under 'Advanced'

Cyberduck

Cyberduck is anonther GUI tool to browse, upload and transfer data from AWS S3/Glacier. When creating a new Bookmark, pick the S3 (with timestamps) service. If it does not show up in the list pick More Options... at the bottom of the list and then search for timestamps in Profiles. This ensures that you can see the original modification date of the uploaded files while browsing and not the date they were uploaded to S3/Glacier. All files called Froster.allfiles.csv are not stored in Glacier but S3 and you can just mark the file and hit the Edit button in the toolbar or run ctrl+k to open that file with the default tool for csv files on your computer (this is often Excel)

More detailed file system analysis

froster index runs pwalk to create a large csv file with all file system information but this file is immedatly deleted after running index and all detailed file information ist lost. You can save that original file (warning: it can be huge) using the --pwalk-copy option and then analyse that data later. You can then use visidata (command: vd) to view the csv file (hit q to quit visidata)

froster index --pwalk-copy ~/my_department.csv /shared/my_department
vd ~/my_department.csv

FAQ and Troubleshooting

Why can't I use Froster to archive to Google Drive, Sharepoint/OneDrive, etc ?

Today Froster only supports S3 compatible stores, but it could be adopted to support more as it is using rclone underneath which supports almost everything. The annoying thing with end user file sharing services such as Drive/OneDrive/Sharepoint is that you need to have an oauth authentication token and this needs to be re-generated often. This is not user friendly on an HPC machine without GUI and web browser and on a standard linux machine it is still not super smooth. Another adjustment needed: you have perhaps seen that the tool creates a tar file for each directory. Currently only files < 1MB are tarred up (this can be changed). At least for Sharepoint one wants to create larger tar archives as the number of total files is limited in Sharepoint. If you are interested in this please submit an issue.

Command line help

Each of the sub commands has a help option, for example froster archive --help

froster --help

dp@grammy:~$ froster --help

usage: froster  [-h] [-c CORES] [-d] [-D] [-i] [-l] [-m MEMORY] [-n] [-p PROFILE] [-v]
                {credentials,crd,config,cnf,index,idx,archive,arc,delete,del,mount,umount,restore,rst,update,upd} ...

A user-friendly archiving tool for teams that move data between high-cost POSIX file systems and low-cost S3-like object storage systems

positional arguments:
  {credentials,crd,config,cnf,index,idx,archive,arc,delete,del,mount,umount,restore,rst,update,upd}
                        sub-command help

options:
  -h, --help            show this help message and exit
  -c CORES, --cores CORES
                        Number of cores to be allocated for the machine. (default=4)
  -d, --debug           verbose output for all commands
  -D, --default-profile
                        Select default profile
  -i, --info            print froster and packages info
  -l, --log-print       Print the log file to the screen
  -m MEMORY, --mem MEMORY
                        Amount of memory to be allocated for the machine in GB. (default=64)
  -n, --no-slurm        do not submit a Slurm job, execute in the foreground.
  -p PROFILE, --profile PROFILE
                        User this profile for the current session
  -v, --version         print froster version

froster config --help

dp@grammy:~$ froster config --help

usage: froster config [-h] [-p] [-r] [-i IMPORT_CONFIG] [-e EXPORT_CONFIG]

Froster configuration bootstrap. This command will guide you through the
configuration of Froster. You can also import and export configurations.

options:
  -h, --help            show this help message and exit
  -p, --print           Print the current configuration
  -r, --reset           Delete the current configuration and start over
  -i IMPORT_CONFIG, --import IMPORT_CONFIG
                        Import a given configuration file
  -e EXPORT_CONFIG, --export EXPORT_CONFIG
                        Export the current configuration to the given directory

froster index --help

dp@grammy:~$ froster index --help

usage: froster index [-h] [-f] [-p] [-y PWALKCOPY] [folders ...]

Scan a file system folder tree using 'pwalk' and generate a hotspots CSV file
that lists the largest folders. As this process is compute intensive the
index job will be automatically submitted to Slurm if the Slurm tools are
found.

positional arguments:
  folders               Folders you would like to index (separated by space), using the pwalk file system crawler 

options:
  -h, --help            show this help message and exit
  -f, --force           Force indexing
  -p, --permissions     Print read and write permissions for the provided folder(s)
  -y PWALKCOPY, --pwalk-copy PWALKCOPY
                        Directory where the pwalk CSV file should be copied to.

froster archive --help

dp@grammy:~$ froster archive --help

usage: froster archive [-h] [-f] [-l LARGER] [-o OLDER] [--newer NEWER] [-n] [-i NIHREF] [-m] [-r] [-s] [-t] [-d] [folders ...]

Select from a list of large folders, that has been created by 'froster index', and
archive a folder to S3/Glacier. Once you select a folder the archive job will be
automatically submitted to Slurm. You can also automate this process

positional arguments:
  folders               folders you would like to archive (separated by space), the last folder in this list is the target   

options:
  -h, --help            show this help message and exit
  -f, --force           Force archiving of a folder that contains the .froster.md5sum file
  -l LARGER, --larger LARGER

                        Archive folders larger than <GiB>. This option
                        works in conjunction with --older <days>. If both
                        options are set froster will print a command that
                        allows you to archive all matching folders at once.
  -o OLDER, --older OLDER

                        Archive folders that have not been accessed more than
                        <days>. (optionally set --mtime to select folders that
                        have not been modified more than <days>). This option
                        works in conjunction with --larger <GiB>. If both
                        options are set froster will print a command that
                        allows you to archive all matching folders at once.
  --newer NEWER, -w NEWER

                        Archive folders that have been accessed within the last 
                        <days>. (optionally set --mtime to select folders that
                        have not been modified more than <days>). This option
                        works in conjunction with --larger <GiB>. If both 
                        options are set froster will print a command that 
                        allows you to archive all matching folders at once.
  -n, --nih             Search and Link Metadata from NIH Reporter
  -i NIHREF, --nih-ref NIHREF
                        Use NIH Reporter reference for the current archive
  -m, --mtime           Use modified file time (mtime) instead of accessed time (atime)
  -r, --recursive       Archive the current folder and all sub-folders
  -s, --reset           
                        This will not download any data, but recusively reset a folder
                        from previous (e.g. failed) archiving attempt.
                        It will delete .froster.md5sum and extract Froster.smallfiles.tar
  -t, --no-tar          Do not move small files to tar file before archiving
  -d, --dry-run         Execute a test archive without actually copying the data

froster delete --help

dp@grammy:~$ froster delete --help

usage: froster delete [-h] [-r] [folders ...]

Remove data from a local filesystem folder that has been confirmed to
be archived (through checksum verification). Use this instead of deleting manually

positional arguments:
  folders          folders (separated by space) from which you would like to delete files, you can only delete files that have been archived

options:
  -h, --help       show this help message and exit
  -r, --recursive  Delete the current archived folder and all archived sub-folders

froster {mount|umount} --help

dp@grammy:~$ froster mount --help
usage: froster mount [-h] [-a] [-l] [-m MOUNTPOINT] [folders ...]

Mount or unmount the remote S3 or Glacier storage in your local file system
at the location of the original folder.

positional arguments:
  folders               archived folders (separated by space) which you would like to mount.

options:
  -h, --help            show this help message and exit
  -a, --aws             Mount folder on new EC2 instance instead of local machine
  -l, --list            List all mounted folders
  -m MOUNTPOINT, --mount-point MOUNTPOINT
                        pick a custom mount point, this only works if you select a single folder.

froster restore --help

fdp@grammy:~$ froster restore --help
usage: froster restore [-h] [-a] [-d DAYS] [-i INSTANCETYPE] [-l] [-m] [-o RETRIEVEOPT] [-r] [folders ...]

Restore data from AWS Glacier to AWS S3 One Zone-IA. You do not need
to download all data to local storage after the restore is complete.
Just use the mount sub command.

positional arguments:
  folders               folders you would like to to restore (separated by space)

options:
  -h, --help            show this help message and exit
  -a, --aws             Restore folder on new AWS EC2 instance instead of local machine
  -d DAYS, --days DAYS  Number of days to keep data in S3 One Zone-IA storage at $10/TiB/month (default: 30)
  -i INSTANCETYPE, --instance-type INSTANCETYPE
                        The EC2 instance type is auto-selected, but you can pick any other type here
  -l, --no-download     skip download to local storage after retrieval from Glacier
  -m, --monitor         Monitor EC2 server for cost and idle time.
  -o RETRIEVEOPT, --retrieve-opt RETRIEVEOPT

                        More information at:
                            https://docs.aws.amazon.com/AmazonS3/latest/userguide/restoring-objects-retrieval-options.html
                            https://aws.amazon.com/es/s3/pricing/

                        S3 GLACIER DEEP ARCHIVE or S3 INTELLIGET-TIERING DEEP ARCHIVE ACCESS
                            Bulk:
                                - Within 48 hours retrieval            <-- default
                                - costs of $2.50 per TiB
                            Standard:
                                - Within 12 hours retrieval
                                - costs of $10 per TiB
                            Expedited:
                                - 9-12 hours retrieval
                                - costs of $30 per TiB

                        S3 GLACIER FLEXIBLE RETRIEVAL or S3 INTELLIGET-TIERING ARCHIVE ACCESS
                            Bulk:
                                - 5-12 hours retrieval
                                - costs of $2.50 per TiB
                            Standard:
                                - 3-5 hours retrieval
                                - costs of $10 per TiB
                            Expedited:
                                - 1-5 minutes retrieval
                                - costs of $30 per TiB

                            In addition to the retrieval cost, AWS will charge you about
                            $10/TiB/month for the duration you keep the data in S3.
                            (Costs in Summer 2023)
  -r, --recursive       Restore the current archived folder and all archived sub-folders

Commercial solutions

You can self-install Froster in a few seconds without requiring root access and you can collaborate well in small teams. However, since Froster requires you having write access to all folders and at least read access to all files you manage, it will not scale to many users. If you are rather looking for a feature rich software managed by IT, you should consider an Enterprise solution such as Starfish. Froster is a good on-ramp to Starfish. If many users in your organization end up using Froster, it is time considering Starfish as an alternative but if fewer than a handful users find Froster useful, you may be able to defer your Starfish project until you have a critical mass. You can access many advanced Starfish features through a web browser while Froster has a simple CLI/TUI interface.

Discontinuing Froster

All good things inevitably come to an end. Consider what you might encounter when attempting to restore your data 15 years from now. While AWS Glacier and Rclone will still exist, we cannot guarantee the continued maintenance of components like Textual or DuckDB, or even Froster itself. However, even if certain tools fade away, you can always rely on utilities like Rclone or Cyberduck (for smaller amounts of data) to retrieve your data, as it is kept in its original format.

Alternatively, the shell script s3-restore.sh simplifies this process, driving Rclone with the appropriate settings. Using the command s3-restore.sh list, you can view all folders archived in the JSON database foster-archives.json (default location: ~/.froster/config/). To restore a specific folder, simply use the command followed by the desired path, for example: s3-restore.sh /my/shared/folder. It's worth noting that system administrators might hesitate to endorse tools written in programming languages they aren't familiar with, such as Python. Fortunately, s3-restore.sh is a straightforward bash shell script, easily customizable to suit specific needs. Note: You may have to change the AWS profile in s3-restore.sh to a profile you find under ~/.aws