diff --git a/docs/assets/images/lifesciences/globus_guest_collection_permissions.png b/docs/assets/images/lifesciences/globus_guest_collection_permissions.png new file mode 100644 index 00000000..c9c5f772 Binary files /dev/null and b/docs/assets/images/lifesciences/globus_guest_collection_permissions.png differ diff --git a/docs/assets/images/lifesciences/globus_transfer_parameter_options.png b/docs/assets/images/lifesciences/globus_transfer_parameter_options.png new file mode 100644 index 00000000..a884b2a1 Binary files /dev/null and b/docs/assets/images/lifesciences/globus_transfer_parameter_options.png differ diff --git a/docs/assets/images/lifesciences/globus_two_pane_view.png b/docs/assets/images/lifesciences/globus_two_pane_view.png new file mode 100644 index 00000000..86ab24dd Binary files /dev/null and b/docs/assets/images/lifesciences/globus_two_pane_view.png differ diff --git a/docs/workshops/genomics_exchange/fall2026/session2/index.md b/docs/workshops/genomics_exchange/fall2026/session2/index.md index a27a38a3..9a4e6ba7 100644 --- a/docs/workshops/genomics_exchange/fall2026/session2/index.md +++ b/docs/workshops/genomics_exchange/fall2026/session2/index.md @@ -12,13 +12,249 @@ tags: - **Date:** Tuesday, September 29, 2026 - **Time:** 11:00 AM – 12:00 PM EDT (10:00 – 11:00 AM CDT) - **Format:** 60 min interactive workshop, online (Microsoft Teams) - - **Instructor:** Arun Seetharam + - **Instructors:** Arun Seetharam and Rose Wilfong (RCAC) - **Register:** [Register for Session 2 on Microsoft Teams](https://events.teams.microsoft.com/event/f99c5474-7bdf-4987-9391-52e14578676b@4130bd39-7c53-419c-b1e5-8758d6d63f21). The Teams join link is sent to registrants. - - **Recording:** TBD - - **Materials:** TBD + - **Recording:** [recording link coming soon] + - **Materials:** This page. - +This page is for Purdue faculty, postdocs, students, and staff in the life sciences who store or analyze research data on RCAC systems, whether or not you attended the session. It explains what each RCAC storage tier (home, scratch, Data Depot, Fortress) is for, how to share data within your lab and archive it to tape, how Globus moves data between tiers and to collaborators, and what to write in an NIH Data Management and Sharing (DMS) Plan. It ends with a five-step checklist to complete before your lab generates new data. No prior HPC experience is needed. An RCAC account and your lab's Data Depot group name help you run the commands. - +## The data lifecycle on RCAC + +RCAC has four storage tiers, each built for a different stage of a project. The most common mistake is leaving the only copy of important data in scratch. + +A typical genomics project moves through the tiers like this: + +1. Raw data arrives from the sequencing core. Put it on **Data Depot** and archive a copy to **Fortress** right away. For sequence data you will share, consider also submitting to NCBI (SRA or GEO) now with a release hold. NCBI keeps it private and releases it when the accession is published. Fortress remains your exact-copy backup; SRA does not return files byte-for-byte as submitted. +2. Copy the inputs your jobs need into **scratch** and run the analysis there. +3. Copy results worth keeping back to **Data Depot**. +4. When the project ends or the paper is published, bundle the project and archive it to **Fortress**. + +**Home** holds your scripts, configuration files, and small documents, never sequencing data. + +| | Home | Scratch | Data Depot | Fortress | +|---|---|---|---|---| +| **Purpose** | Scripts, configs, small files | Working space for running jobs | Shared, active lab data | Long-term archive (tape) | +| **Capacity** | 25 GB per user (TODO(arun): verify) | Large per-user quota; check with `myquota` | Purchased in 1 TB increments; 100 GB free trial | No quota | +| **Backup** | Nightly snapshots, kept up to about 3 months | None | Nightly snapshots, kept up to about 3 months; mirrored at two campus sites | Two copies on separate media; no protection against deletion | +| **Cost** | Included | Included | See the [RCAC orders page](https://www.rcac.purdue.edu/orders/products?category=3) (sign-in required) | Free with RCAC cluster access | +| **Purge** | Never | Files not accessed or modified in 60 days (30 days on Bell and Anvil) | Never | Never | +| **Who pays** | Included with your RCAC account | Included with your RCAC account | The lab (PI), annually | Included with RCAC services; storing more than 1 PB may incur a cost recovery charge | + +!!! warning "Scratch purge" + Scratch is not backed up, and RCAC no longer sends warning emails before a purge. Run `purgelist` on a cluster to see which of your scratch files are scheduled for removal. + +Snapshots let you recover from accidental deletion for a limited time, but they are not a backup. A file deleted on the day it was created cannot be recovered. See [Data Depot lost file recovery](../../../../userguides/depot/recover/index.md). + +## Depot and Fortress in practice + +### Getting Data Depot for your lab + +- The PI purchases capacity on the [Data Depot Purchase](https://www.rcac.purdue.edu/purchase) page, in 1 TB increments at an annual rate. You do not need to own cluster nodes. +- To try it first, [request a free 100 GB trial](https://www.rcac.purdue.edu/orders/products?category=3). +- Data on Depot belongs to the PI's research group, not to individuals, so it stays with the lab when students and postdocs leave. +- Each Depot group also gets a shared Fortress group space at `/group/mylab/`. + +### Sharing and permissions + +A lab's Depot space is `/depot/mylab/`, where `mylab` is your group name. By default it has a `data/` folder for shared research data and an `apps/` folder for shared software. Unix groups control access: for example, write access to `/depot/mylab/data/` requires membership in `mylab-data`. The PI or a designee manages membership on the [RCAC group management page](https://www.rcac.purdue.edu/account/groups). New members must log out and back in before their access takes effect. + +Depot is suitable for non-HIPAA human subjects data. It is not approved for HIPAA, ePHI, FISMA, ITAR, or other regulated data. + +### Fortress: hsi and htar + +Fortress is a tape library. You cannot log in to it with SSH. From any RCAC cluster, use `hsi` to manage files and `htar` to create archives. No keytab setup is needed on RCAC systems. + +Fortress stores a few large files efficiently but many small files very slowly. A file under about 30 to 50 MB counts as "small" at Fortress scale. Bundle directories of FASTQ files, per-sample outputs, or pipeline work folders into archives before sending them to Fortress. `htar` does this in one step, without needing local disk space for the archive. + +### Commands + +Replace `mylab` with your group name and `project_2026` with your project directory. + +Check which lab storage groups you belong to: + +```bash +groups +``` + +Count the files in a project directory before archiving (thousands of files means bundle them): + +```bash +find /depot/mylab/data/project_2026 -type f | wc -l +``` + +Bundle a project directory into one archive in your lab's Fortress space, with checksum verification: + +```bash +cd /depot/mylab/data +htar -Hverify=1 -cvf /group/mylab/project_2026.tar project_2026 +``` + +List the contents of that archive without restoring it: + +```bash +htar -tvf /group/mylab/project_2026.tar +``` + +List what is in your lab's Fortress space: + +```bash +hsi ls -l /group/mylab +``` + +!!! note + `htar` cannot archive a single file larger than 64 GB. Use `htar_large` for those files, or compress and `hsi put` them individually. Archiving a large project can take hours, so run it inside a batch job rather than on a login node. + +!!! danger "Deletion on Fortress is permanent" + Fortress keeps two copies of every file to protect against media failure, but if you delete or overwrite a file on Fortress, it cannot be recovered. + +## Globus +Globus is a managed file-transfer service designed for moving large research datasets reliably and efficiently. It is especially useful for transfers that involve many files, hundreds of gigabytes or more, long transfer times, or data moving between institutions. Rather than keeping a terminal session or browser download running, you can submit the transfer to Globus and let the service manage it for you. + +### Why use Globus for large transfers? + +A long-running `scp` or `sftp` command can be interrupted by a dropped network connection, a closed laptop, or an expired terminal session. The user may then need to determine what arrived and restart or resume the transfer manually. Downloading to a laptop and uploading again is even less efficient when the real goal is to move data between two remote storage systems. + +Globus uses a **fire-and-forget** model: + +1. You select the source, destination, and files, then submit one transfer request. +2. Globus transfers the data directly between the two collections. Your browser is only the control interface and is not in the data path. +3. You may close the browser or sign out. Globus continues to monitor the task, tunes the transfer for performance, and retries recoverable network or system failures. When possible, it resumes from the point of failure rather than starting the entire transfer again. +4. Globus verifies file integrity with checksums and records the task in the **Activity** tab. It can email you when the transfer succeeds or when a problem needs your attention. + +This makes Globus especially useful for transfers measured in hundreds of gigabytes or terabytes: users can start the work and return to their research instead of keeping a terminal open and watching the connection. “Fire-and-forget” does not mean “never check”. Users should confirm that the task reports **Succeeded** before deleting the source copy. Problems such as an expired login, insufficient destination space, or missing permissions still require user action; after the problem is corrected, Globus can continue the task. + +### Globus endpoints at RCAC + +In the Globus interface, a named location you can browse is called a **collection**. RCAC provides collections for most clusters and storage systems. Each collection is a doorway to a particular storage system; it does not create a new copy of the data. + +To find a Purdue collection, select a collection search bar in File Manager and enter `Purdue` plus the cluster or storage-system name. Choose the collection carefully because some clusters expose home and scratch storage together, while others use separate collections. + +#### Cluster home and scratch storage + +For **Anvil**, **Negishi**, and **Gautschi**, home and scratch storage are available through one collection: + +- **Home and scratch:** `Purdue {Name} Cluster`, replacing `{Name}` with the cluster name. For example, search for `Purdue Anvil Cluster`. After opening the collection, use the path field and directory browser to move between the home and scratch filesystems. + +For **Bell** and **Gilbreth**, home and scratch storage use separate collections: + +- **Home directories:** `Purdue {Name} Cluster - Home Directories` +- **Scratch directories:** `Purdue {Name} Cluster - Scratch` + +For example, a Gilbreth user should select `Purdue Gilbreth Cluster - Home Directories` for files in home and `Purdue Gilbreth Cluster - Scratch` for files in scratch. If the expected files are not visible, first confirm that you opened the correct collection. + +#### Group and archival storage + +These collections are independent of the cluster collections: + +- **Research Data Depot:** `Purdue Research Computing - Data Depot`. Search for `Purdue Data Depot`, then browse to your group's directory, such as `/depot/mylab/data/`. +- **Fortress:** `Purdue Fortress HPSS Archive`. Search for `Purdue Fortress`, then browse to your personal or group archive space, such as `/group/mylab/`. + +Opening a collection does not grant additional access. Globus uses your RCAC identity and the underlying Unix permissions, so you will see only the directories and files your account is authorized to use. If a lab member cannot open a Depot group directory, verify their group membership rather than creating a new collection. + +### Transferring to Depot and Fortress + +1. Navigate to the [RCAC Globus transfer portal](https://transfer.rcac.purdue.edu/). +2. Sign in with your Purdue account. On your first visit, approve the prompts that connect your Purdue identity to Globus. If you have not recently authorized your credentials, you will be prompted to log in with MFA. +3. Open **File Manager** and switch to the two-panel view using the *Panels* options. Choose a collection for each panel. + + +