Install a Multi-Node Deployment

This section presents instructions on deploying Autonomous Identity in a multi-node deployment. Multi-node deployments are configured in production environments, providing performant throughput by distributing the processing load across servers and supporting failover redundancy.

Like single-node deployment, ForgeRock provides a deployer script that pulls a Docker image from ForgeRock’s Google Cloud Registry (gcr.io) repository. The image contains the microservices, analytics, and backend databases needed for the system. The deployer also uses the node IP addresses specified in your hosts file to set up an overlay network and your nodes.

The topology presented in this section is a generalized example that is used in our automated testing. Each production deployment is unique and requires proper review prior to implementation.

For production, the example assumes that you run the deployer on a dedicated low-spec box. After you set up your environment and back up the autoid-config directory, you can recycle the deployer box.

Figure 8: An example multi-node deployment.

Autonomous Identity deployed in a multi-node deployment

Prerequisites

Let’s deploy Autonomous Identity on a multi-node target on CentOS 7. The following are prerequisites:

Operating System. The target machine requires CentOS 7. The deployer machine can use any operating system as long as Docker is installed. For this chapter, we use CentOS 7 as its base operating system.

Default Shell. The default shell for the autoid user must be bash.

Subnet Requirements. We recommend deploying your multi-node machines within the same subnet. Ports must be open for the installation to succeed. Each instance should be able to communicate to the other instances.

If any hosts used for the Docker cluster (docker-managers, docker-workers) have an IP address in the range of 10.0.x.x, they will conflict with the Swarm network. As a result, the services in the cluster will not connect to the Cassandra database or Elasticsearch backend.

The Docker cluster hosts must be in a subnet that provides IP addresses 10.10.1.x or higher.

Deployment Requirements. Autonomous Identity provides a deployer.sh script that downloads and installs the necessary Docker images. To download the deployment images, you must first obtain a registry key to log into the ForgeRock Google Cloud Registry (gcr.io). The registry key is only available to ForgeRock Autonomous Identity customers. For specific instructions on obtaining the registry key, see How To Configure Service Credentials (Push Auth, Docker) in Backstage.
Filesystem Requirements. Autonomous Identity requires a shared filesystem accessible from the Spark master, Spark worker, analytics hosts, and application layer. The shared filesystem should be mounted at the same mount directory on all of those hosts. If the mount directory for the shared filesystem is different from the default, /data , update the /autoid-config/vars.yml file to point to the correct directories:
```
analytics_data_dir: /data
analytics_conf_dif: /data/conf
```
Architecture Requirements. Make sure that the Spark master is on a separate node from the Spark workers.
Database Requirements. Decide which database you are using: Apache Cassandra or MongoDB. The configuration procedure is slightly different for each database.
Deployment Best-Practice. The example combines the ODFE data and Kibana nodes. For best performance in production, dedicate a separate node to Elasticsearch, data nodes, and Kibana.
IPv4 Forwarding. Many high-security environments run their CentOS-based systems with IPv4 forwarding disabled. However, Docker Swarm does not work with a disabled IPv4 forward setting. In such environments, make sure to enable IPv4 forwarding in the file /etc/sysctl.conf:
```
net.ipv4.ip_forward=1
```

We recommend that your deployer team have someone with Cassandra expertise. This guide is not sufficient to troubleshoot any issues that may arise.

Example Topology

Make sure you have sufficient storage for your particular deployment. For more information on sizing considerations, see Deployment Planning Guide.

Each deployment is unique and should be discussed with your installer and ForgeRock.

For this example, use the following configuration for this example multi-node deployment:

Suggested Topology
	Num Nodes	Cores	Memory
Deployer	1	2 vCPU	4 GB
Docker Manager	1	8 vCPU	32 GB
Docker Worker	1	8 vCPU	32 GB
Cassandra Seeds	2	8 vCPU	32 GB
Spark Master	1	16 vCPU	64 GB
Spark Worker	2	8 vCPU	32 GB
Elasticsearch (ODFE) Master/Kibana	1	8 vCPU	32 GB
ODFE Data	1	8 vCPU	32 GB

Set Up the Nodes

Set up your VMs based on the Example Topology.

For each VM, make sure that you have CentOS 7 as your operating system. Check your CentOS 7 version.
```
$ sudo cat /etc/centos-release
```
Set the user for the target node to a username of your choice:
1. In this example, create user autoid.
  $ sudo adduser autoid
2. Set the password for the user you created in the previous step.
  $ sudo passwd autoid
3. Configure the user for passwordless sudo.
  $ echo "autoid ALL=(ALL) NOPASSWD:ALL" | sudo tee /etc/sudoers.d/autoid
4. Add administrator privileges to the user.
  $ sudo usermod -aG wheel autoid
Change to the user account.
```
$ su - autoid
```
Install yum-utils package on the deployer machine. yum-utils is a utilities manager for the Yum RPM package repository. The repository compresses software packages for Linux distributions.
```
$ sudo yum install -y yum-utils
```

Install Docker on the Deployer Machine

Install Docker on the deployer machine. We run commands from this machine to install Autonomous Identity on the target machine. In this example, we use CentOS 7.

Change to the user account.
```
$ su - autoid
```
Create the installation directory. Note that you can use any install directory for your system as long as your run the deployer.sh script from there. Also, the disk volume where you have the install directory must have at least 8GB free space for the installation.
```
$ mkdir autoid-config
```

Set up the Docker-CE repository.

$ sudo yum-config-manager \
     --add-repo https://download.docker.com/linux/centos/docker-ce.repo

Install the latest version of the Docker CE, the command-line interface, and containerd.io, a containerized website.
```
$ sudo yum install -y docker-ce docker-ce-cli containerd.io
```
Enable Docker to start at boot.
```
$ sudo systemctl enable docker
```
Start Docker.
```
$ sudo systemctl start docker
```
Check that Docker is running.
```
$ systemctl status docker
```
Add the user to the Docker group.
```
$ sudo usermod -aG docker ${USER}
```
Logout of the user account.
```
$ logout
```
Re-login using created user. Login with the user created for the deployer machine. For example, autoid.
```
$ su - autoid
```

Set Up SSH on the Deployer

On the deployer machine, change to the ~/.ssh directory.
```
$ cd ~/.ssh
```
Run ssh-keygen to generate an RSA keypair, and then click Enter. You can use the default filename.

Do not add a key passphrase as it results in a build error.
```
$ ssh-keygen -t rsa -C "autoid"
```
The public and private rsa key pair is stored in home-directory/.ssh/id_rsa and home-directory/.ssh/id_rsa.pub .
Copy the SSH key to the autoid-config directory.
```
$ cp id_rsa ~/autoid-config
```
Change the privileges to the file.
```
$ chmod 400 ~/autoid-config/id_rsa
```
Copy your public SSH key, id_rsa.pub, to each of your nodes.

If your target system does not have an ~/.ssh/authorized_keys, create it using sudo mkdir -p ~/.ssh, then sudo touch ~/.ssh/authorized_keys.

For this example, copy the SSH key to each node:
```
$ ssh-copy-id -i id_rsa.pub autoid@<Node IP Address>
```
On the deployer machine, test your SSH connection to each target machine. This is a critical step. Make sure the connection works before proceeding with the installation.

For example, SSH to first node:
```
$ ssh -i id_rsa autoid@<Node 1 IP Address>

Last login: Sat Oct 3 03:02:40 2020
```
If you can successfully SSH to each machine, set the privileges on your ~/.ssh and ~/.ssh/authorized_keys.
```
$ chmod 700 ~/.ssh && chmod 600 ~/.ssh/authorized_keys
```
Enter Exit to end your SSH session.
Repeat steps 5–8 again for each node.

Set Up a Shared Directory

The analytics master and worker nodes require a shared directory, typically, /data. There are numerous ways to set up a shared directory, the following procedure is just one example and sets up an NFS server on the analytics master.

On the Analytics Spark Master node, install nfs-utils. This step may require that you run the install with root privileges, such as sudo or equivalent.
```
$ sudo yum install -y nfs-utils
```
Create the /data directory.
```
$ mkdir -p /data
```

Change the permissions on the /data directory.

$ chmod -R 755 /data
$ chown nfsnobody:nfsnobody /data

Start the services and enable them to start at boot.

$ systemctl enable rpcbind
$ systemctl enable nfs-server
$ systemctl enable nfs-lock
$ systemctl enable nfs-idmap

$ systemctl start rpcbind
$ systemctl start nfs-server
$ systemctl start nfs-lock
$ systemctl start nfs-idmap

Define the sharing points in the /etc/exports file.
```
$ vi /etc/exports

/data  <Remote IP Address 1>(rw,sync,no_root_squash,no_all_squash)
/data  <Remote IP Address 2>(rw,sync,no_root_squash,no_all_squash)
```
If you change the domain name and target environment, you need to also change the certificates to reflect the new changes. For more information, see Customize Domains.
Start the NFS service.
```
$ systemctl restart nfs-server
```

Add the NFS service to the firewall-cmd public zone service:

$ firewall-cmd --permanent --zone=public --add-service=nfs
$ firewall-cmd --permanent --zone=public --add-service=mountd
$ firewall-cmd --permanent --zone=public --add-service=rpc-bind
$ firewall-cmd --reload

On each spark worker node, run the following:
1. Install nfs-utils:
  $ yum install -y nfs-utils
2. Create the NFS directory mount points:
  $ mkdir -p /data
3. Mount the NFS shared directory:
  $ mount -t nfs <NFS Server IP>:/data /data
4. Test the new shared directory by creating a small text file. On an analytics worker node, run the following, and then check for the presence of the test file on the other servers:
  $ cd /data $ touch test

Change Network Kernal Settings for Cassandra

The default network kernal setttings require overriding the default values to ensure TCP buffers are properly sized for use with Cassandra.

If you are running an evaluation deployment with a small dataset, you can skip this section. For production deployments using Cassandra, run these instructions.

For each Cassandra seed node, run the following steps:

SSH to a Cassandra seed machine.
Change to the default user. In this example, autoid.

Open a text editor and edit the /etc/sysctl.conf file. Add the following settings to the file:

net.core.rmem_max=16777216
net.core.wmem_max=16777216
net.core.rmem_default=16777216
net.core.wmem_default=16777216
net.core.optmem_max=40960
net.ipv4.tcp_rmem=4096 87380 16777216
net.ipv4.tcp_wmem=4096 65536 16777216
vm.max_map_count=1048575

At runtime, make sure swap is not used. For a permanent change, modify /etc/fstab file.
```
$ sudo swapoff -a
```
Disable defrag of huge pages. The following command must be executed at each boot:
```
$ echo never | sudo tee /sys/kernel/mm/transparent_hugepage/defrag
```
Set the proper ulimits for the user running Cassandra by setting them in the /etc/security/limits.conf file. In this example, the user running Cassandra is autoid.

Open a text editor, and add the following settings in the /etc/security/limits.conf file:
```
autoid - memlock unlimited
autoid - nofile 100000
autoid - nproc 32768
autoid - as unlimited
```

Install Autonomous Identity

Before you begin, make sure you have CentOS 7 installed on your target machine.

On the deployer machine, change to the installation directory.
```
$ cd ~/autoid-config/
```
Obtain the registry key for the ForgeRock Google Cloud Registry (gcr.io). The registry key is only available to ForgeRock Autonomous Identity customers. For specific instructions on obtaining the registry key, see How To Configure Service Credentials (Push Auth, Docker) in Backstage.

$ docker login -u _json_key -p "$(cat autoid_registry_key.json)" https://gcr.io/forgerock-autoid

You should see:

Login Succeeded

Run the create-template command to generate the deployer.sh script wrapper. Note that the command sets the configuration directory on the target node to /config. Note that the --user parameter eliminates the need to use sudo while editing the hosts file and other configuration files.
```
$ docker run --user=$(id -u) -v ~/autoid-config:/config -it gcr.io/forgerock-autoid/deployer:2021.8.2 create-template
```

Configure Autonomous Identity

The create-template command from the previous section creates a number of configuration files, required for the deployment: ansible.cfg, vars.yml, vault.yml, and hosts.

On the deployer node, change to the autoid-config/ directory.
Check the ansible.cfg file for the remote user and SSH private key file location. If you followed these instruction, the default settings should be in place, which will not require any edits to the ansible.cfg file.
```
[defaults]
host_key_checking = False
remote_user = autoid
private_key_file = id_rsa
```
On the deployer machine, open a text editor and edit the ~/autoid-config/vars.yml file to configure specific settings for your deployment:
- AI Product. Do not change this property.
  ai_product: auto-id
- Domain and Target Environment. Set the domain name and target environment specific to your deployment by editing the /autoid-config/vars.xml file. By default, the domain name is set to forgerock.com and the target environment is set to autoid. The default Autonomous Identity URL will be: https://autoid-ui.forgerock.com. For this example, we use the default values.
  domain_name: forgerock.com target_environment: autoid
  If you change the domain name and target environment, you need to also change the certificates to reflect the new changes. For more information, see Customize Domains.
- Analytics Data Directory and Analytics Configuration Direction. For a multi-node Spark deployment, Autonomous Identity requires a shared filesystem accessible from Spark Master, Spark Worker(s), and Analytics hosts. The shared filesystem should be mounted at same mount directory on all of the above hosts. If the mount directory for shared filesystem is different than /data, update the following properties in the vars.yaml file to point to the correct location:
  analytics_data_dir: /data analytics_conf_dif: /data/conf
- Database Type. By default, Apache Cassandra is set as the default database for Autonomous Identity. For MongoDB, set the db_driver_type: to mongo.
  db_driver_type: cassandra
- Private IP Address Mapping. You can skip this step as we use the private IP addresses in the subnet.
- Authentication Option. This property has three possible values:
  - Local. Local indicates that sets up elasticsearch with local accounts and enables the Autonomous Identity UI features: self-service and manage identities. Local auth mode should be enabled for demo environments only.
  - SSO. SSO indicates that single sign-on (SSO) is in use. With SSO only, the Autonomous Identity UI features, self-service and manage identities pages, is not available on the system but is managed by the SSO provider. The login page displays "Sign in using OpenID." For more information, see Set Up SSO.
  - LocalAndSSO. LocalAndSSO indicates that SSO is used and local account features, like self-service and manage identities are available to the user. The login page displays "Sign in using OpenID" and a link "Or sign in via email".
    
    authentication_option: "Local"
- Access Log. By default, the access log is enabled. If you want to disable the access log, set the access_log_enabled variable to "false".
- JWT Expiry and Secret File. Optional. By default, the session JWT is set at 30 minutes. To change this value, set the jwt_expiry property to a different value.
  jwt_expiry: "30 minutes" jwt_secret_file: "{{install path}}"/jwt/secret.txt" jwt_audience: "http://my.service" oidc_jwks_url: "na"
- Local Auth Mode Password. When authentication_option is set to Local, the local_auth_mode_password sets the password for the login user.
- SSO. Use these properties to set up SSO. For more information, see Set Up SSO.
- MongoDB Configuration. For MongoDB clusters, enable replication by uncommenting the mongodb_replication_replset property.
  # uncomment below for mongo with replication enabled. Not needed for single node deployments mongodb_replication_replset: mongors
  Also, enable a custom key for inter-machine authentication in the clustered nodes.
  # custom key # password for inter-process authentication # please regenerate this file on production environment with # command 'openssl rand -base64 741' mongodb_keyfile_content: | 8pYcxvCqoe89kcp33KuTtKVf5MoHGEFjTnudrq5BosvWRoIxLowmdjrmUpVfAivh CHjqM6w0zVBytAxH1lW+7teMYe6eDn2S/O/1YlRRiW57bWU3zjliW3VdguJar5i Z+1a8lI+0S9pWynbv9+Ao0aXFjSJYVxAm/w7DJbVRGcPhsPmExiSBDw8szfQ8PAU 2hwRl7nqPZZMMR+uQThg/zV9rOzHJmkqZtsO4UJSilG9euLCYrzW2hdoPuCrEDhu Vsi5+nwAgYR9dP2oWkmGN1dwRe0ixSIM2UzFgpaXZaMOG6VztmFrlVXh8oFDRGM0 cGrFHcnGF7oUGfWnI2Cekngk64dHA2qD7WxXPbQ/svn9EfTY5aPw5lXzKA87Ds8p KHVFUYvmA6wVsxb/riGLwc+XZlb6M9gqHn1XSpsnYRjF6UzfRcRR2WyCxLZELaqu iKxLKB5FYqMBH7Sqg3qBCtE53vZ7T1nefq5RFzmykviYP63Uhu/A2EQatrMnaFPl TTG5CaPjob45CBSyMrheYRWKqxdWN93BTgiTW7p0U6RB0/OCUbsVX6IG3I9N8Uqt l8Kc+7aOmtUqFkwo8w30prIOjStMrokxNsuK9KTUiPu2cj7gwYQ574vV3hQvQPAr hhb9ohKr0zoPQt31iTj0FDkJzPepeuzqeq8F51HB56RZKpXdRTfY8G6OaOT68cV5 vP1O6T/okFKrl41FQ3CyYN5eRHyRTK99zTytrjoP2EbtIZ18z+bg/angRHYNzbgk lc3jpiGzs1ZWHD0nxOmHCMhU4usEcFbV6FlOxzlwrsEhHkeiununlCsNHatiDgzp ZWLnP/mXKV992/Jhu0Z577DHlh+3JIYx0PceB9yzACJ8MNARHF7QpBkhtuGMGZpF T+c73exupZFxItXs1Bnhe3djgE3MKKyYvxNUIbcTJoe7nhVMrwO/7lBSpVLvC4p3 wR700U0LDaGGQpslGtiE56SemgoP
  On production deployments, you can regenerate this file by running the following command:
  $ openssl rand -base64 741
- Elasticsearch Heap Size. The default heap size for Elasticsearch is 1GB, which is too small for production. For production deployments for large datasets, uncomment the option and enter a value.
  #elastic_heap_size: 4g # sets the heap size (1g|2g|3g) for the Elastic Servers
- Java API Service. Optional. Set the Java API Service (JAS) properties for the deployment: authentication, maximum memory, directory for attribute mappings data source entities:
  jas: auth_enabled: true auth_type: 'jwt' signiture_key_id: 'service1-hmac' signiture_algorithm: 'hmac-sha256' max_memory: 4096M mapping_entity_type: /common/mappings datasource_entity_type: /common/datasources

Open a text editor and enter the private IP addresses of the target machines in the ~/autoid-config/hosts file. Make sure the target host IP addresses are accessible from the deployer machine. The following is an example of the ~/autoid-config/hosts file. NOTE: [notebook] is not used in Autonomous Identity.

Click to See a Host File for a Multi-Node Cassandra Deployment

If you configured Cassandra as your database, the ~/autoid-config/hosts file is as follows for multi-node target deployments:

[docker-managers]
10.128.0.90

[docker-workers]
10.128.0.170

[docker:children]
docker-managers
docker-workers

[cassandra-seeds]
10.128.0.175
10.128.0.34

[spark-master]
10.128.0.180

[spark-workers]
10.128.0.176
10.128.0.177

[spark:children]
spark-master
spark-workers

[mongo_master]
#ip#  mongodb_master=True

[mongo_replicas]
#ip-1#
##ip-2#
##…

[mongo:children]
mongo_replicas
mongo_master

# ELastic Nodes
[odfe-master-node]
10.128.0.178

[odfe-data-nodes]
10.128.0.184

[kibana-node]
10.128.0.184

[notebook]
#ip#

Open a text editor and set the Autonomous Identity passwords for the configuration service, elasticsearch backend, and Cassandra or MongoDB database. The vault passwords file is located at ~/autoid-config/vault.yml.

Despite the presence of special characters in the examples below, do not include special characters, such as & or $, in your production vault.yml passwords as it will result in a failed deployer process.

configuration_service_vault:
    basic_auth_password: ~@C~O>@%^()-_+=|<Y*$$rH&&/m#g{?-o!z/1}2??3=!*&

cassandra_vault:
    cassandra_password: ~@C~O>@%^()-_+=|<Y*$$rH&&/m#g{?-o!z/1}2??3=!*&
    cassandra_admin_password: ~@C~O>@%^()-_+=|<Y*$$rH&&/m#g{?-o!z/1}2??3=!*&
    keystore_password: Acc#1234
    truststore_password: Acc#1234

mongo_vault:
    mongo_admin_password: ~@C~O>@%^()-_+=|<Y*$$rH&&/m#g{?-o!z/1}2??3=!*&
    mongo_root_password: ~@C~O>@%^()-_+=|<Y*$$rH&&/m#g{?-o!z/1}2??3=!*&
    keystore_password: Acc#1234
    truststore_password: Acc#1234

elastic_vault:
    elastic_admin_password: ~@C~O>@%^()-_+=|<Y*$$rH&&/m#g{?-o!z/1}2??3=!*&
    elasticsearch_password: ~@C~O>@%^()-_+=|<Y*$$rH&&/m#g{?-o!z/1}2??3=!*&
    keystore_password: Acc#1234
    truststore_password: Acc#1234

Encrypt the vault file that stores the Autonomous Identity passwords, located at ~/autoid-config/vault.yml. The encrypted passwords will be saved to /config/.autoid_vault_password . The /config/ mount is internal to the deployer container.
```
$ ./deployer.sh encrypt-vault
```
Download the images. This step downloads software dependencies needed for the deployment and places them in the autoid-packages directory.
```
$ ./deployer.sh download-images
```
Run the deployment.
```
$ ./deployer.sh run
```

Set the Replication Factor

Once Cassandra has been deployed, you need to set the replication factor to match the number of nodes on your system. This ensures that each record is stored in each of the nodes. In the event one node is lost, the remaining node can continue to serve content even if the cluster itself is running with reduced redundancy.

You can define replication on a per keyspace-basis as follows:

SSH to a Cassandra seed node.
Change to the /opt/autoid/apache-cassandra-3.11.2/ directory.

Start the Cassandra sheel, cqlsh, and define the autoid keyspace. Change the replication factor to match the number of seed nodes. The default admin user for Cassandra is zoran_dba.

$ bin/cqlsh -u zoran_dba

$ zoran_dba@cqlsh> desc keyspace autoid;
CREATE KEYSPACE autoid WITH replication = {'class':'SimpleStrategy','replication_factor':'2'} AND durable_writes=true;

CREATE TABLE autoid.user_access_decisions_history(
  user text,
  entitlement text,
  date_created timestamp,
  …

Restart Cassandra on this node.
Repeat these steps on the other Cassandra seed node(s).

Resolve Hostname

After installing Autonomous Identity, set up the hostname resolution for your deployment.

Resolve the hostname:

Configure your DNS servers to access Autonomous Identity dashboard on the target node. The following domain names must resolve to the IP address of the target node: <target-environment>-ui.<domain-name>.
If DNS cannot resolve target node hostname, edit it locally on the machine that you want to access Autonomous Identity using a browser. Open a text editor and add an entry in the /etc/hosts (Linux/Unix) file or C:\Windows\System32\drivers\etc\hosts (Windows) for the self-service and UI services for each managed target node.
```
<Target IP Address>  <target-environment>-ui.<domain-name>
```
For example:
```
34.70.190.144  autoid-ui.forgerock.com
```
If you set up a custom domain name and target environment, add the entries in /etc/hosts. For example:
```
34.70.190.144  myid-ui.abc.com
```
For more information on customizing your domain name, see Customize Domains.

Access the Dashboard

Access the Autonomous Identity console UI:

Open a browser. If you set up your own url, use it for your login.
```
$ https://autoid-ui.forgerock.com/
```

test user: bob.rodgers@forgerock.com
password: <password>

Check Apache Cassandra

Check Cassandra:

On the target node, check the status of Apache Cassandra.

$ /opt/autoid/apache-cassandra-3.11.2/bin/nodetool status

An example output is as follows:

Datacenter: datacenter1
=======================
Status=Up/Down
|/ State=Normal/Leaving/Joining/Moving
--  Address       Load       Tokens       Owns (effective)  Host ID                               Rack
UN  34.70.190.144 1.33 MiB   256          100.0%            a10a91a4-96e83dd-85a2-4f90d19224d9  rack1
--

Check MongoDB

Check the status of MongoDB:

On the target node, check the status of MongoDB.

$ mongo --tls \
--host <Host IP> \
--tlsCAFile /opt/autoid/mongo/certs/rootCA.pem  \
--tlsAllowInvalidCertificates  \
--tlsCertificateKeyFile /opt/autoid/mongo/certs/mongodb.pem

Check Apache Spark

Check Spark:

SSH to the target node and open Spark dashboard using the bundled text-mode web browser
```
$ elinks http://localhost:8080
```
You should see Spark Master status as ALIVE and worker(s) with State ALIVE.

Click to See an Example of the Spark Dashboard

Start the Analytics

If the previous installation steps all succeeded, you must now prepare your data’s entity definitions, data sources, and attribute mappings prior to running your analytics jobs. These step are required and are critical for a successful analytics process.

For more information, see Set Entity Definitions.

Autonomous Identity 2021.8.2