Globus Connect Server Installation Guide

1. Introduction

This installation guide provides an overview of Globus Connect Server v5 for system administrators who will install and operate the service.

Globus Connect Server v5 is the next evolution of the server software. It provides new capabilities and enhancements for both administrators and users, and platform features to build interesting solutions for data management.

The latest version, 5.4, supports the following features:

  • Deployments with multiple data transfer nodes

  • Guest collections (data sharing with collaborators)

  • Mapped collections (access for users with local accounts)

  • HTTPS access to data - for direct access from browsers and other HTTPS clients

  • GridFTP access to data - for reliable, bulk data transfer via the Globus transfer service

  • High assurance features for management of protected data

  • Support for the following storage systems: ActiveScale, Azure Blob Storage, S3, Box, Ceph, Dropbox, Google Drive, Google Cloud Storage, HPSS, iRODS, OneDrive, POSIX, POSIX with file staging, SpectraLogic BlackPearl

  • An integrated OIDC server for local authentication using any service configured via PAM

2. Globus Connect Version 5 Terminology

The Globus Connect Server architecture has evolved to support several new capabilities. This section provides an overview of the components in Globus Connect Server version 5 and how they relate to version 4 components.

  • Endpoint (changed from version 4): The endpoint is a deployment of Globus Connect Server version 5. A single endpoint may optionally include multiple Data Transfer Nodes (DTNs) or servers. Each server or DTN is referred to as Node. The endpoint provides the interface for server management and configuration.

  • Storage connector: A storage connector allows the endpoint to use a particular type of storage. (E.g., POSIX file system, Google Drive.) You may configure multiple storage connectors for a single endpoint, allowing simultaneous access to all connectors.

  • Storage gateway: Storage gateways provide the storage access policies for the endpoint’s connected storage systems. A storage gateway is a named, discoverable interface by which authorized users can create and manage collections on a connected storage system. A connected storage system may have multiple storage gateways.

  • Collection: Collections provide the data access interfaces, allowing access via HTTPS (client/server access), GridFTP (asynchronous bulk transfer), and REST API (for advanced operations). In Globus Connect Server version 5, a collection is a named set of files (or blobs), hierarchically organized in folders, associated with a specific storage gateway. Access to a collection is authenticated with Globus Auth-issued OAuth2 access tokens, with data access policies defined in the collection itself. Globus Connect Server version 5 supports two types of collections:

    • Mapped collection: Each user accessing the collection must have a local account on the storage system. Their Globus identity is mapped to their local account. In version 4, these are called "host endpoints."

    • Guest collection: Users can access the collection without a local account on the storage system. Access is based on permissions granted by an authorized user via Globus. In version 4, these are called "shared endpoints."

Globus Connect Server version 5

With the above architecture, Globus Connect Server version 5 supports many new features including:

  • Multiple storage types connected to the same endpoint

  • Multiple storage gateways against the same storage type

  • Clear separation between management and configuration, and data access interfaces

  • End-to-end backup and synchronization of configuration data on endpoints

  • HTTPS access to the data in addition to bulk data access via GridFTP.

3. Prerequisites

Important
The prerequisites listed in this section must be met before you begin to install Globus Connect Server v5 on your system. Contact us if you have any questions regarding the prerequisites.

3.1. Supported Linux distributions

Globus Connect Server version 5 is currently supported on the following Linux x86-64 distributions:

  • Red Hat Enterprise Linux 8, 9

  • Rocky Linux 8, 9

  • AlmaLinux 8, 9

  • Springdale Linux 8, 9

  • Oracle Linux 8, 9

  • CentOS 9 Stream

  • Fedora 40, 41

  • Debian 11, Debian 12

  • Ubuntu 20.04 LTS, 22.04 LTS, 24.04 LTS, 24.10

  • SUSE Linux Enterprise Server 15.5, 15.6

  • OpenSUSE Leap 15.5, 15.6

Note
Support and update package availablilty generally ends at the official EOL date of a distribution.

3.2. Memory

The minimum memory requirement for Globus Connect Server v5 is 2GB of RAM.

Important
While GCS can run in an environment with these minimum specifications, this is not a suggested configuration for a production environment and should not be used to evaluate performance.

3.3. Administrator privileges

You must have administrator (root) privileges on your system to install Globus Connect Server version 5; sudo can be used to perform the installation.

3.4. System time synchronization

Your system must be running ntpd or another daemon for synchronizing with standard time servers.

3.5. Unicode Locale

Your system must use a unicode-capable locale in order to run the Globus Connect Server command-line tools. For RedHat, CentOS, and Fedora systems, you can use the en_US.UTF-8 locale, and for Debian and Ubuntu systems, you can use the C.UTF-8 locale.

3.6. Internet-accessible system

Other hosts on the Internet must be able to initiate connections to the system where you will be installing Globus Connect Server v5. Your network administrator may be able to offer assistance if you run into problems, or contact us.

3.7. Open TCP ports

If your system is behind a firewall, several TCP ports must be opened for Globus to work. You may need to coordinate with your network or security administrator to open the ports.

The TCP ports that must be open for the default Globus Connect Server v5 installation are as follows.

  • Ports 50000—​51000 inbound and outbound to/from ANY

    • Used for GridFTP data channel traffic.

    • Ports in this range will be bound and used by GridFTP only during transfers as needed.

    • The use of the default port range is strongly recommended (you can read why here).

    • Data channel traffic is sent directly between endpoints—​it is not relayed by the Globus service.

  • Port 443 inbound from ANY

    • Used by Globus Connect Server v5 Manager Service

    • Used for GridFTP control channel traffic.

    • Used for HTTPS access to collections.

  • Port 443 outbound to ANY

    • Used to communicate with the Globus service via its REST API.

    • Used to communicate with cloud storage services.

    • Used to pull Globus Connect Server v5 packages from the Globus repository.

3.7.1. Changing defaults

The above are for the default installation of Globus Connect Server. Some of the ports can be changed via the globus-connect-server command-line tool.

Data channel traffic

The incoming and outgoing port ranges can be changed on a per-node basis by using the globus-connect-server node update command with the --incoming-port-range or --outgoing-port-range command-line options. Use of the default port range is strongly recommended as discussed in our Understanding Data Channel Traffic doc.

Control Channel Incoming Traffic

The control channel port can be changed on a per-endpoint basis by using the globus-connect-server endpoint update command with the --gridftp-control-channel-port command-line option. This is available in Globus Connect Server version 5.4.55 or higher.

Note
The --outgoing-port-range defines the source port used for outbound data channel connections to other GridFTP servers. By default, this is unconstrained and will use the next available port in the servers ephemeral port range.

3.8. NAT Support

NAT deployments are supported provided that the ports needed by Globus Connect Server are properly mapped from the server’s public IP address to the server’s internal IP address and provided that systems in the NATed environment can connect to the server via its' public IP address (NAT loopback or similar).

When deploying a new node in your endpoint, you’ll need to use the --ip-address option for the globus-connect-server node setup command to ensure that your server’s public IP address gets registered for the node rather than the server’s internal IP address.

In scenarios where systems in the NATed environment cannot connect to the server via its' public IP address there will be certain caveats to consider:

  • You’ll need to use the --use-explicit-host option to force the GCS CLI commands you run on your server to connect to the GCS Manager on the local host. If you don’t use this option, then the GCS CLI will attempt to connect to your node’s public IP address.

  • Transfers between multiple collections on the same endpoint, or between multiple endpoints in the NATed environment, may require advanced configuration and will not provide the same user experience. Contact support if you believe you are in this situation.

  • User client systems in the NATed environment may not be able to connect to the GCS Manager, causing limited functionality for such users in terms of creating guest collections and managing credentials. Resolving such issues may require advanced configuration and may not provide the same user experience. Contact support if you believe you are in this situation.

4. Installation

A Globus Connect Server v5 deployment includes installation of multiple components: (a) endpoint for servers and networking configuration, (b) storage gateways with connectors for storage system and policies information, and (c) mapped collections for users to access data. If you are familiar with installation of Globus Connect Server version 4, some of your assumptions about the installation process may no longer be valid. Please follow these instructions carefully.

This section covers the installation and setup of endpoint, and links to next sections for storage gateway and collection setup. As we walk through each part of this installation, links to alternate configurations and connectors (e.g., Google Drive storage gateway) will be provided. You can customize and fine-tune this configuration to your specific needs later without doing a reinstallation. Please see the Globus Connect Server version 5 Quickstart Guide for basic installation on a single node for access to a POSIX filesystem.

Before continuing, please confirm that the prerequisites detailed in the previous section have been met.

4.1. Install Globus Connect Server v5 software

Skip to the appropriate section for your Linux distribution and follow the instructions to install Globus Connect Server v5 on your system.

This must be done on each system which will be acting as a Data Transfer Node for the endpoint(s) you create.

sudo dnf install https://dl.fedoraproject.org/pub/epel/epel-release-latest-8.noarch.rpm
sudo dnf install https://downloads.globus.org/globus-connect-server/stable/installers/repo/rpm/globus-repo-latest.noarch.rpm

Install the DNF config manager:

sudo dnf install 'dnf-command(config-manager)'

Install Globus Connect Server:

sudo dnf install globus-connect-server54
sudo dnf install https://dl.fedoraproject.org/pub/epel/epel-release-latest-9.noarch.rpm
sudo dnf install https://downloads.globus.org/globus-connect-server/stable/installers/repo/rpm/globus-repo-latest.noarch.rpm

Install the DNF config manager:

sudo dnf install 'dnf-command(config-manager)'

Install Globus Connect Server:

sudo dnf install globus-connect-server54
sudo dnf install https://downloads.globus.org/globus-connect-server/stable/installers/repo/rpm/globus-repo-latest.noarch.rpm

install Globus Connect Server:

sudo dnf install globus-connect-server54
curl -LOs https://downloads.globus.org/globus-connect-server/stable/installers/repo/deb/globus-repo_latest_all.deb
sudo dpkg -i globus-repo_latest_all.deb
sudo apt update
sudo apt install globus-connect-server54
curl -LOs https://downloads.globus.org/globus-connect-server/stable/installers/repo/deb/globus-repo_latest_all.deb
sudo dpkg -i globus-repo_latest_all.deb
sudo apt update
sudo apt install globus-connect-server54

If you are using SUSE Linux Enterprise server, you must enable the PackageHub repository to install all of the Globus Connect Server dependencies.

sudo rpm --import https://downloads.globus.org/globus-connect-server/stable/installers/keys/GPG-KEY-Globus
sudo zypper install https://downloads.globus.org/globus-connect-server/stable/installers/repo/rpm/globus-repo-latest.noarch.rpm
sudo zypper install globus-connect-server54

4.2. Create the endpoint

With the Globus Connect Server v5 software installed on your server, the next step is to create the endpoint. You will run the globus-connect-server endpoint setup command to setup the endpoint across all of the Globus services it will need to function.

Note
Prior to Globus Connect Server version 5.4.61, the endpoint was manually registered with Globus Auth by logging into the Globus Developers web interface. That step provided the client ID and secret needed for globus-connect-server endpoint setup and other commands. This step is now automated within the globus-connect-server endpoint setup command; newer endpoints will no longer need to manage these credentials. Previous endpoints should continue to supply the credentials to CLI commands as describe below and in the CLI reference.

To create the endpoint, run the globus-connect-server endpoint setup command. This command need not be run as root user, and is run only once per endpoint, on the first node (server/DTN) setup for the endpoint. Subsequent servers/nodes that are added should not use this command, and separate instructions are provided for that.

This command creates an endpoint in the Globus Transfer service and obtains a DNS name and certificate for hosting the endpoint. In addition, it generates an encryption key used to confidentially manage configuration data which is backed up to Globus hosted service for management. The encryption key is written to the file deployment-key.json in the current directory. The data in this file is necessary to configure Data Transfer Nodes to host Globus Connect services.

Important
The deployment-key.json file contains encryption keys to ensure configuration data backed up is encrypted. In the case of endpoints with multiple DTNs, this key will be needed on every Data Transfer Node, and cannot be recovered by Globus.

The globus-connect-server endpoint setup command takes a number of command-line options, but the following ones are required to create an endpoint:

DISPLAY_NAME

Name for the Endpoint

--organization string

Organization operating the Endpoint

--owner string

Identity username of the user who will be the administrator of this endpoint. (e.g. janedoe@example.edu). This must be an identity that has logged into Globus.

--contact-email ADDRESS

Email address of the support contact for this endpoint.

Globus Connect Server prior to version 5.4.61 required manual registration with Globus Auth via the Developers web interface, and specifying client credentials to the GCS CLI required the --client-id option. If you are running a GCS version prior to 5.4.61 or your GCS deployment key was created prior to version 5.4.61, you must use the --client-id option to this command.

Important
Globus Connect Server clients and their projects can be viewed and managed in the Globus Developers web interface (https://app.globus.org/settings/developers). Do not delete the endpoint client from the Developers web interface; it must exist for the lifetime of the endpoint.

Once run, the command prompts you to agree to Let’s Encrypt’s terms of service for having Globus Connect Server obtain a certificate for the endpoint using Lets Encrypt’s ACME protocol. The command takes a few minutes to complete as certificates are provisioned on the endpoint. Finally, it will provide a link to a login page so that the endpoint owner can set the advertised owner of the endpoint in the Globus Transfer service.

Example
globus-connect-server endpoint setup "My GCSv5.4 Endpoint" \
    --organization "Example Organization" \
    --owner admin@example.edu \
    --contact-email support@example.edu

The command returns information about the endpoint that may be useful for additional configuration later, including the domain name of the endpoint, a link to send to subscription managers to set the endpoint as managed, and the redirect URI needed if Google Drive or Cloud connectors will be used with this endpoint.

The command will register the new endpoint with Globus Auth by creating a new Auth client under an existing Globus Auth project. If you do not have an existing Globus Auth project, one will be automatically created for you with the value of --owner as the project admin. If you have more than one project, the command will print the IDs of available projects. In that case, rerun the command using the --project-id option.

See globus-connect-server endpoint setup documentation for more information about all options to this command including additional information about the endpoint such as description and contact information.

Note
If you set up an endpoint and want to set the endpoint owner to an account to which you do not have access, you can use the --dont-set-advertised-owner option. The endpoint will still be initialized, but the advertised owner string won’t be set in the Transfer service. Users will not be able to discover the owner of the endpoint when searching in the Globus Transfer service. You can change this (after a node is set up) by having the endpoint owner use the command globus-connect-server endpoint set-owner-string to update the endpoint.

4.3. Set up services on the Data Transfer Node

Run the globus-connect-server node setup command to configure and start the Globus services on the Data Transfer Node. This command must be done as the root user, as it enables and starts systemd services. The deployment-key.json file from the previous step will be used by this command.

Globus Connect Server prior to version 5.4.61 required manual registration with Globus Auth via the Developers web interface, and specifying client credentials to the GCS CLI required the --client-id option. If you are running a GCS version prior to 5.4.61 or your GCS deployment key was created prior to version 5.4.61, you must use the --client-id option to this command.

Important
If you are deploying Globus Connect Server on a diskless server, see Diskless Data Transfer Node Management for information on how to configure Globus Connect Server to avoid problems.
Example
sudo globus-connect-server node setup

See globus-connect-server node setup for more information about all options to this command.

4.4. Add Data Transfer Nodes to the endpoint

Endpoints can have multiple Data Transfer Nodes and these are called "nodes". For each Data Transfer Node, the Globus Connect Server software must be downloaded and installed as described in install section.

Once installation of the software is complete, the node will be set up to be included in the endpoint. The setup requires the deployment-key.json file from the previous step so the file should be copied over to each Data Transfer Node.

Run the globus-connect-server node setup command to configure and start the Globus services on each Data Transfer Node. This command must be done as the root user as it enables and starts systemd services. The deployment-key.json file from the previous step must be available on the Data Transfer Nodes, and the command must be run once on each Data Transfer Node to enable the Globus services on that node.

Example
sudo globus-connect-server node setup

See globus-connect-server node setup for more information about all options to this command.

4.5. Log into the endpoint

For further configuration and management, tokens for authentication to the Globus Connect Server Manager service are needed. The globus-connect-server login command can be used to log into the endpoint for further configuration. This command will print out a login URL to follow to authenticate with Globus, and return an access code that needs to be pasted into the command-line tool. This will result in an authentication token that can be used to talk to the Globus Connect Server management API. As long as you are on a Data Transfer Node of an endpoint the "localhost" argument can be used to get credentials for that endpoint.

Login Command Example
globus-connect-server login localhost

See globus-connect-server login for more information about all options to this command.

4.6. View endpoint configuration

You can now use the globus-connect-server endpoint show command to display the configuration of the endpoint.

Example
globus-connect-server endpoint show

Display Name:    My GCSv5.4 Endpoint
ID:              a44ca3ef-a8b9-4b73-aa5f-546ed5ab7e66
Subscription ID: None
Public:          True
GCS Manager URL: https://09ad6.0.glob.us
Network Use:     normal
Organization:    Example Organization

4.7. Set the endpoint as managed

Endpoints that require premium functionality—​such as guest collections for data sharing and premium connectors—​must be managed under a Globus subscription.

If your organization has a subscription, and your Globus account has the subscription manager role, you may set the endpoint as managed using the globus-connect-server command as follows.

Set the Endpoint as Managed
globus-connect-server endpoint set-subscription-id DEFAULT

If you are not the subscription manager for your organization, you will have to request your organization’s subscription manager to set the endpoint as managed. They can refer to the FAQ on using the Globus Web App to set that option.

4.8. Assign roles

Users or groups can be granted various roles on any managed endpoint, with each role granting the user or group different privileges with respect to that endpoint. Endpoint roles can be managed via the globus-connect-server endpoint role command or via the endpoints page of the Globus management console.

For example, to add an activity manager role, you could do the following:

Example
globus-connect-server endpoint role create activity_manager manager@example.org
Role ID: e9edfc32-d3b6-45e3-b970-3c68af56859d

See the Endpoint Role command-line documentation for information on what roles are available, and details about how to manage roles.

5. Next Steps: Create a Gateway and Mapped Collection

At this point, you’ve installed Globus Connect Server v5 on your Data Transfer Nodes and created an endpoint. However, an endpoint does not support data access without a storage gateway and mapped collection configured. You should follow the Data Access Guide to configure gateways and collections on your endpoint so users can access data via Globus.

Read the Command Line Reference for a complete description of the globus-connect-server command-line tool.

6. Generating and Monitoring Log Files

6.1. GridFTP Server Log Files

By default, the GridFTP log is located at:

/var/log/gridftp.log

The configuration settings for the GridFTP log file are found in this file:

/etc/gridftp.d/globus-connect-server

Logging for the GridFTP service is enabled by default. Additional details concerning logging for the GridFTP server are available in the globus-gridftp-server man page here.

A logrotate config is installed to rotate the log weekly, retaining the last 5 logs.

Note
Modifications to /etc/gridftp.d/globus-connect-server* are not allowed and will be overwritten. If it is necessary to add custom GridFTP configurations, admins should create new files in /etc/gridftp.d/, avoiding filenames beginning with globus*. Keep in mind that configuration files are processed in ASCII order (eg. a config. file named z_logging would be processed after globus-connect-server, and would override any settings set earlier).

6.2. High Assurance Audit Log Files

Endpoints that have a high assurance subscription will have high assurance audit logs available at:

/var/log/gridftp-audit.log

The audit log contains a record for each login and file access on a high assurance collection. The format is explained in the audit log format reference. The globus-connect-server audit commands can be used for searching or exporting logs in standard formats.

A logrotate config is installed to rotate the log daily, retaining the last 90 logs to the /var/log/gridftp-audit/ directory.

Filenames and directory paths from high assurance gateways may contain sensitive data. GridFTP and HTTP server logs contain filenames and directory paths, and logs from high assurance gateways should be managed appropriately.

6.3. HTTPS Collection Access Logs

Apache logs for HTTPS collection access and errors are logged to the default Apache log path for your platform (/var/log/httpd or /var/log/apache2).

Starting with Globus Connect Server version 5.4.82, HTTPS collection access is logged to the file globus_access_log with the following format, which is similar to the Apache combined format with additional collection information.

<collection hostname>:<collection id> <client address> <authenticated globus username> <time> <request> <response status> <size> <client referrer> <client user agent>

Example:

m-019423.fe55.data.globus.org:bb31dd6f-cb77-4ff6-a487-d7bc6b545632 53.176.55.222 user@example.org [06/Feb/2025:19:49:04 +0000] "GET /data/example.dat?download HTTP/1.1" 200 188593 "-" "Mozilla/5.0 (Windows NT 10.0; Win64; x64; rv:135.0) Gecko/20100101 Firefox/135.0"

6.4. Globus Connect Server Manager API Log

The Globus Connect Server Manager API Log is stored in

/var/log/globus-connect-server/gcs-manager/gcs.log

7. Globus Help Resources

7.1. Documentation Website

This website (docs.globus.org) contains a wealth of information about configuring and using the Globus service. Many common issues can be resolved quickly by browsing our frequently asked questions and reading the relevant guides and how-tos. We recommend consulting these resources first when looking for fast resolution to any issue you are having with the Globus service.

7.2. Mailing Lists

If you use Globus, then participating in one or more of the public email lists is an excellent way to keep in touch with your peers in the Globus Community. For questions about managing your Globus deployment, e.g. installing software for a Globus endpoint, configuring your firewall, and integrating your institution’s identity system, subscribe to the admin list. For other inquiries and discussions, try the user or developer lists. For more information on mailing lists and how to subscribe, click here.

7.3. Globus Support

Questions or issues that pertain to Globus Connect Server v5 installation or to any client or service that is used in the Globus software-as-a-service (SaaS) or platform-as-a-service (PaaS) offering can be directed to the Globus support team by submitting a ticket. Subscriptions include a guaranteed support service level.

When submitting a ticket for an issue with Globus Connect Server, please include the endpoint name, a description of your issue, and screenshot/text dumps of any errors you are seeing. Please also include the output of Globus Connect Server’s self-diagnostic command, run as root, from the server hosting the endpoint:

globus-connect-server self-diagnostic

8. Upgrading Globus Connect Server

Use the following instructions to update your installation. During the process of updating the packages, the Globus services on your Data Transfer Node will be restarted.

Note
If you are upgrading from a version of GCSv5 prior to 5.4.67, you must also execute the GCS CLI Endpoint Key Convert command in order to update your Endpoints Deployment Key (ie. deployment-key.json) and make it compatible with the current release of GCS CLI.

We have further details and documentation on this update available here.

8.1. Single Node Endpoint 

$ sudo yum repository-packages Globus-Connect-Server-5-Stable upgrade
$ sudo dnf repository-packages Globus-Connect-Server-5-Stable upgrade
$ sudo apt-get update
$ sudo apt-get install --only-upgrade ".*globus.*"
$ sudo dnf repository-packages Globus-Connect-Server-5-Stable upgrade
$ sudo apt-get update
$ sudo apt-get install --only-upgrade ".*globus.*"
sudo zypper update -r Globus-Connect-Server-5-Stable update '*globus*'

If the node belongs to an existing endpoint, run this command after all nodes in the endpoint have been upgraded in order to upgrade GCS policies. This command only needs to be run once for the existing endpoint.

$ sudo globus-connect-server endpoint upgrade

8.2. Multiple Node Endpoint

Important

If you have multiple Data Transfer Nodes serving the same endpoint, then you must run the same version of the Globus Connect Server software on all of them. If you do not, there may be problems synchronizing the configuration between nodes and the service may not function correctly on all nodes.

If your endpoint consists of more than a single node, you can perform rolling updates in order to minimize downtime for the endpoint. By following the instructions below, one or more nodes can service user data transfers while the remaining nodes are upgraded and then the upgraded nodes can be brought into service while the other nodes are upgraded. Keep in mind the important note above, it is important that all enabled nodes in the endpoint are on the same version of Globus Connect Server.

  • Select 1 or more nodes from the endpoint to perform the initial upgrade. Selecting half of the nodes is preferred in order to maintain data transfer throughput.

  • Disable the nodes selected for the first round of Globus Connect Server upgrades.

    $ globus-connect-server node update --disable e6968b77-266d-45c4-9d2b-a65a3a2b4fee
Message: Updated node e6968b77-266d-45c4-9d2b-a65a3a2b4fee

  • Follow the single node upgrade steps for each node in the set of disabled nodes.

  • Once the first nodes are upgraded, disable the active nodes in the endpoint. Do not enable the nodes in the first group until after this is done as you do not want to be running different versions of Globus Connect Server on various nodes in an endpoint.

  • Enable one of the upgraded nodes by running the node setup command on it locally.

    Globus Connect Server prior to version 5.4.61 required manual registration with Globus Auth via the Developers web interface, and specifying client credentials to the GCS CLI required the --client-id option. If you are running a GCS version prior to 5.4.61 or your GCS deployment key was created prior to version 5.4.61, you must use the --client-id option to this command.

    $ sudo globus-connect-server node setup

  • Enable the rest of the upgraded nodes.

    $ globus-connect-server node update --enable e6968b77-266d-45c4-9d2b-a65a3a2b4fee
Message: Updated node e6968b77-266d-45c4-9d2b-a65a3a2b4fee

  • Upgrade GCS on the nodes in the second group using the instructions in single node upgrade.

  • Enable the nodes in the second group.

Appendix A: Diskless Data Transfer Node Management

If you plan on using Globus Connect Server v5 with diskless Data Transfer Nodes, you can use the following commands to prepare a Data Transfer Node node configuration that you can use to bootstrap the node setup command while retaining the node id and node-specific data encryption key. This will prevent with Data Transfer Node from creating a new encryption key at each boot (which will eventually cause problems), and ensure that when a node is deployed on a new IP address the Globus DNS record will be updated if the Data Transfer Node IP address changes.

A.1. Create the Data Transfer Node Configuration

Use the command globus-connect-server node create to export the configuration to use on a Data Transfer Node. This command can be run on any machine which has Globus Connect Server installed on it, and does not need to be run on the Data Transfer Node on which the configuration will be used. This command generates a new node ID and encryption key, and exports it to a JSON file which can be consumed by globus-connect-server node setup's --import-node command-line option.

The globus-connect-server node create command was added in Globus Connect Server v5.4.6

Repeat this once for each Data Transfer Node you will be configuring, using the --export-node command-line option to choose the output file path.

Prepare Three Data Transfer Node configurations
globus-connect-server node create --export-node node1.json
globus-connect-server node create --export-node node2.json
globus-connect-server node create --export-node node3.json

A.2. Deploy the Data Transfer Node Configurations

At Data Transfer Node initialization, you can use a command-line like the following to set up the node and start the services. This example assumes that the file /config/gcs.sh defines shell variables $GLOBUS_DEPLOYMENT_KEY to be the path to the deployment key and $NODE_CONFIG as the path to one of the node-specific configuration files created in the previous step.

Globus Connect Server prior to version 5.4.61 required manual registration with Globus Auth via the Developers web interface, and specifying client credentials to the GCS CLI required the --client-id option. If you are running a GCS version prior to 5.4.61 or your GCS deployment key was created prior to version 5.4.61, you must use the --client-id option to this command.

Setup Node Reusing the Node ID and Encryption Key
. /config/gcs.sh

globus-connect-server node setup \
    --deployment-key "$GLOBUS_DEPLOYMENT_KEY" \
    --import-node "$NODE_CONFIG"

Appendix B: Understanding Data Channel Traffic

The data channel is where Globus Connect Server actually transmits the data that is being moved between endpoints. The default port range used for data channel connections is TCP 50000 to 51000. We strongly recommend that all endpoints be configured to use the default data port range, as this will provide maximum compatibility with other endpoints that are also configured to use the default data port range and have their firewall rules configured to allow traffic in this range. If your endpoint uses a non-default data port range, then you are - in effect - requiring other sites to potentially have to create additional firewall rules in order to be able to communicate properly with your endpoint. Many sites will not want to do this, which will thus limit the ability of your endpoint to interoperate with the majority of endpoints which are configured to use the default port range.

If two endpoints (ep1 and ep2) are to be able to successfully conduct transfers, then those endpoints must each be able to connect to each other in their configured data port ranges. For example, consider the following:

Globus Connect Server ep1 uses data port range 40000 to 41000

Globus Connect Server ep2 uses data port range 50000 to 51000

When two Globus Connect Server endpoints attempt to conduct a transfer, the endpoint that will be the recipient in that transfer picks out a port (or ports) in its configured data port range that it will listen on to receive the transfer from the sender endpoint. This port value gets communicated back from the receiver endpoint to the sender endpoint via GridFTP control channel data mediated by the Globus service, which both the sender and recipient are listening to on port 443 by default. Once the sender endpoint receives the data port range info for the recipient endpoint, it then initiates an outbound connection to that port (or ports) on the recipient to conduct the actual data transfer. The recipient endpoint will only listen on the selected data port(s) for the duration of the transfer. Once the transfer is over, the recipient endpoint will no longer bind the chosen data port nor listen on it.

To illustrate, consider the case of ep1 and ep2 mentioned above. If ep1 wanted to send ep2 a file, then ep2 would pick out a port (or ports) in its configured data port range of 50000 to 51000. For the sake of example let’s say that port 50021 has been chosen. This value would then get communicated from ep2 to ep1, via the Globus service through the GridFTP control channel that both ep1 and ep2 are listening to. At that point, ep1 would then initiate a connection out to port 50021 on ep2.

To further illustrate, consider again the case of ep1 and ep2 mentioned above. If ep2 wanted to send ep1 a file, then ep1 would pick out a port (or ports) in its configured data port range of 40000 to 41000. For the sake of example let’s say that port 40331 has been chosen. This value would then get communicated from ep1 to ep2, via the Globus service through the GridFTP control channel that both ep1 and ep2 are listening to. At that point, ep2 would then initiate a connection out to port 40331 on ep1.

It is also important to consider what happens in cases where one endpoint is a Globus Connect Server endpoint and the other endpoint is a Globus Connect Personal endpoint. In such cases, the Globus Connect Personal endpoint will always initiate the connection to the Globus Connect Server endpoint for the transfer. Thus, it will always be the Globus Connect Server endpoint that picks the port (or ports) on which it will listen for that connection. This is the case irrespective of which endpoint is the sender or the recipient. As discussed previously, this information gets communicated from the Globus Connect Server endpoint to the Globus Connect Personal endpoint via the Globus service.

After looking at the example given we can see that, in terms of firewall rules, the outbound rules for ep1 must allow it to connect outbound to ep2 on ep2’s configured data port range if ep1 is to be able to send files to ep2. In terms of inbound rules, the firewall rules for ep1 must be configured to allow it to accept inbound connections on its own configured data port range for it to be able to receive files from other endpoints. The firewall rules for the data port range of any endpoint will be similar, and must allow outbound connections to the configured data port range of a remote endpoint for the local endpoint to be able to send files to the remote endpoint, and must allow inbound connections to the configured data port range of the local endpoint for that endpoint to be able to receive files from other endpoints.

As illustrated, an endpoint must be able to receive inbound connections on its own configured data port range, as well as be able to make outbound connections to the data port range of any endpoint it wishes to communicate with. If all Globus Connect Server admins pick their own custom port ranges, then this quickly leads to a situation in which site firewall policies become littered with custom rules for these various port ranges and endpoints. However, if everyone uses the default data port range, then firewall rules are much more predictable and manageable. It is for this reason that we recommend that everyone use the default data port range for their endpoint. Those who use a custom data port range may find that they have problems with their endpoint being able to communicate with other endpoints, for the reasons detailed above. Those using custom data port ranges may also find that the admins of other sites and endpoints may not be willing to set up custom firewall rules to accommodate custom data port range choices.

Appendix C: Decommissioning an Endpoint

The following sections explain how to tear down a Globus Connect Server v5 endpoint and return its nodes to pre Globus Connect Server v5 installation condition. Care should be taken to perform these steps in order to properly decommission the endpoint.

C.1. Removing Nodes from the Endpoint

The following command will remove the node from the endpoint configuration and disable all Globus Connect Server services running on the local node. This command must be run from the node which is being removed.

sudo globus-connect-server node cleanup

This command should be repeated on all nodes in the endpoint before deleting the endpoint. See node cleanup in the CLI reference for more details on the command usage.

C.2. Deleting the Endpoint

The following command will delete the endpoint configuration in all Globus services. This process is irreversible, be sure that all nodes in the endpoint have been removed and that the endpoint is no longer needed before proceeding.

Globus Connect Server prior to version 5.4.61 required manual registration with Globus Auth via the Developers web interface, and specifying client credentials to the GCS CLI required the --client-id option. If you are running a GCS version prior to 5.4.61 or your GCS deployment key was created prior to version 5.4.61, you must use the --client-id option to this command.

globus-connect-server endpoint cleanup \
    --deployment-key deployment-key.json

C.3. Deleting the Client Registration

Note
As of version 5.4.67, deleting the endpoint’s client registration is performed automatically when running endpoint cleanup without the --skip-delete-registration option.

This is the final step in deletion of the endpoint. Before proceeding, make sure that all nodes in the endpoint have been removed and that the endpoint has already been deleted. This step is irreversible.

  1. Log into the Globus Developers web interface, app.globus.org/settings/developers, using the identity used when creating the endpoint.

  2. Select the project on the right that contains the endpoint registration.

  3. Find the Globus Connect Server client definition which contains the endpoint credentials.

  4. Click Delete App Registration.

C.4. Removing Globus Software from a Node

Skip to the appropriate section for your Linux distribution and follow the instructions to remove Globus Connect Server v5 from your system.

sudo yum remove \*globus*\
sudo dnf remove \*globus\*
sudo apt purge \*globus\*

Run the following command on the node after removing the Globus Connect Server v5 software in order to delete any remaining Globus Connect Server files and directories. Many of these files may not be present on your node after successful uninstallation of the Globus Connect Server v5 software and may likely only be present if your configuration included custom modifications.

sudo rm -rf /etc/globus/ \
            /etc/grid-security/ \
            /etc/gridftp.d/ \
            /etc/gridftp.conf \
            /etc/apt/preferences.d/globus \
            /etc/logrotate.d/globus-connect-server \
            /var/lib/globusoidc/ \
            /var/lib/globus-connect-server/ \
            /var/log/gridftp.log* \
            /var/log/globus-connect-server/ \
            /run/globus_oidc.sock \
            /run/globus-connect-server

Appendix D: Network Use Options

Globus Connect Server’s network use options are available for subscribed endpoints. Administrators can configure the amount of parallelism and concurrency that are available to the Globus Transfer service when accessing data on an endpoint’s collections. Administrators can set the network use options on an endpoint or on a storage gateway. The value set on an endpoint acts as a default for all collections except those on storage gateways which have their own setting.

Network use is set to the "Normal" level by default. Endpoints that have multiple physical servers with high-performance network and storage can set higher levels to ensure that Globus uses the bandwidth available, while smaller deployments can set this to lower levels.

Configuration settings from both the source and destination collections are used to determine the concurrency and parallelism options used for a given transfer. Transfers can then use a higher value if both collections support it, but Globus will not overwhelm smaller-capacity endpoints during transfers with larger-capacity endpoints.

Three preset options are provided for the endpoint administrator, which have the following values:

Option Value

Minimal

MaxConcurrency = 1

PreferredConcurrency = 1

MaxParallelism = 1

PreferredParallelism = 1

Normal (Default)

MaxConcurrency = number of servers * 4

PreferredConcurrency = number of servers * 2

MaxParallelism = 8

PreferredParallelism = 4

Aggressive

MaxConcurrency = number of servers * 8

PreferredConcurrency = number of servers * 4

MaxParallelism = 16

PreferredParallelism = 4

In addition to above, an administrator can choose the "Custom" option that lets them set absolute values for both concurrency and parallelism. All these options have a limit of 64 for MaximumConcurrency and MaximumParallelism. These values can be modified by using the --network-use option on the globus-connect-server endpoint update or the globus-connect-server storage-gateway update CONNECTOR commands.

For a given transfer, the concurrency is calculated as the smallest value across the MaximumConcurrency values of both endpoints, and the maximum of the PreferredConcurrency of both endpoints. Parallelism is also calculated similarly, with an additional consideration for transfers with high latency (trans-oceanic transfers) where the parallelism is set to minimum of the Maximum Parallelism value set for both endpoints.

Appendix E: Advanced Apache Configuration

When you run globus-connect-server node setup it enables the Apache web server and some modules it requires, as well as enabling virtual host configuration for access to the GCS Manager API, HTTPS access to collections (if enabled) and the Globus OIDC (if configured).

If your data transfer node is already using the Apache web server, the GCS configuration should not conflict with that. However you may run into problems if you are running another service on TCP port 443. See the Proxy Configuration Guide for information about how to configure the Apache web server to proxy other web services.

Also, most operating systems provide a landing page as part of their default Apache configuration. This landing page is probably unwanted, and may be secured with a self-signed certificate. This is often noticed by network scanners as a configuration error. See the Default VirtualHost Configuration Guide for information about how to update the default configuration to use a valid certificate, restrict the pages it serves, or disable it altogether.