fix docs

docs/2vm.md

@@ -0,0 +1,389 @@
+# TL;DR All-in-one Quick Start Guide
+
+To finish this guide, you will need TWO machines (or VMs) far away from each other. 
+One acts as the Cloud, and the other acts as the Edge. You will need to identify the IP address of the Cloud server.
+
+The following instructions have been tested using AWS EC2 t2.micro with Ubuntu 22.04 LTS and `starlight v0.3.2`.
+
+`git checkout v0.6.2`
+
+---
+
+## The "Cloud"
+
+In this machine you will need to set up the Starlight Proxy and a standard container registry. 
+If you are using AWS EC2, please add the following ports to the Security Group whitelist when you create the VM:
+ - TCP 8090: Starlight Proxy (set the source to be your VPC)
+ - TCP 8080: Adminer - for Metadata database  (set the source to be your own IP)
+ - TCP 5000: Container Registry  (set the source to be your VPC)
+ - TCP 80: Container Registry (set the source to be your VPC)
+
+
+1. Change the hostname of the server (Don't copy and paste, replace `<ip address>` with your server's IP address)
+
+    ```shell
+    echo "cloud" | sudo tee /etc/hostname > /dev/null
+    sudo hostname -F /etc/hostname
+    echo "<ip address> cloud.cluster.local" | sudo tee -a /etc/hosts > /dev/null
+    ```
+
+2. Install [Docker](https://docs.docker.com/engine/install/ubuntu/#install-using-the-repository) and [Docker Compose](https://docs.docker.com/compose/install/)  
+
+    If using Ubuntu 22.04 LTS, you could install Docker and Docker Compose using the following commands: 
+    ```shell
+    sudo apt update && \
+    sudo apt upgrade -y && \
+    sudo apt install -y docker-compose && \
+    sudo usermod -aG docker $USER
+    ```
+    After adding the current user to the `docker` group, you (may) **need to log out and log in** to take effect.
+    To confirm that Docker is working with correct permission, `docker ps` should not print any errors.
+    ```shell
+    docker ps
+    # CONTAINER ID   IMAGE     COMMAND   CREATED   STATUS    PORTS     NAMES
+    ```
+
+3. Clone this project and launch the registry and proxy containers from `./demo/compose/registry+proxy`
+
+    ```shell
+    git clone https://github.com/mc256/starlight.git && \
+    cd starlight && \
+    git checkout v0.6.2 && \
+    cd demo/compose/ && \
+    cp docker-compose-example.yaml docker-compose.yaml && \
+    docker-compose up -d
+    # Creating network "registryproxy_default" with the default driver
+    # Creating registryproxy_db_1       ... done
+    # Creating registryproxy_registry_1 ... done
+    # Creating registryproxy_dbadmin_1  ... done
+    # Creating registryproxy_proxy_1    ... done
+    ```
+    The Starlight proxy writes image metadata to the Postgres database, and
+    the container registry saves container images to `./data_registry`.
+
+
+3. Verify the registry and proxy are running.
+    ```shell
+    # This checks the Starlight Proxy
+    curl http://cloud.cluster.local:8090/
+    # {"status":"OK","code":200,"message":"Starlight Proxy"}
+    # This checks the container registry
+    curl http://cloud.cluster.local:5000/v2/
+    # {}
+    ```
+
+    If it does not work, please restart the containers after the database has been created (missing a db health check).
+    We could put a Nginx reverse proxy to handle SSL certificates or load balancing.
+    But for simplicity, this part is ignored in this example.
+
+
+3. Adjust the TCP window size (Optional). 
+   If the edge node is far away, we will need to adjust the TCP window size so that the connection can speed up to the speed limit faster. (You could calculate the best TCP window size using https://www.speedguide.net/bdp.php later)
+   If you skip this step, the connection will be much slower (the impact on speed is worse for other methods!).
+
+    ```shell
+    cat <<EOT | sudo tee -a /etc/sysctl.conf > /dev/null
+    net.core.wmem_max=125829120
+    net.core.rmem_max=125829120
+    net.ipv4.tcp_rmem= 10240 87380 125829120
+    net.ipv4.tcp_wmem= 10240 87380 125829120
+    net.ipv4.tcp_window_scaling = 1
+    net.ipv4.tcp_timestamps = 1
+    net.ipv4.tcp_sack = 1
+    net.ipv4.tcp_no_metrics_save = 1
+    net.core.netdev_max_backlog = 10000
+    EOT
+    sudo sysctl -p
+    ```
+
+🙌 That's it. Please obtain the IP address of this machine.
+
+---
+
+## The "Edge"
+
+Please get another machine (or VM), you will need to set up a container worker with Starlight Snapshotter plugin.
+
+### 1. Install Dependencies
+
+The worker machine needs `build-essential` and `containerd`.
+```shell
+sudo apt update && sudo apt upgrade -y && \
+sudo apt install -y build-essential containerd
+```
+
+Enable `containerd`
+```shell
+sudo systemctl enable containerd  && \
+sudo systemctl start containerd
+```
+
+Verify `containerd` is running
+```shell
+sudo systemctl status containerd
+#      Active: active
+```
+
+Install Go https://go.dev/doc/install ➡️
+```shell
+wget https://go.dev/dl/go1.20.8.linux-amd64.tar.gz && \
+sudo rm -rf /usr/local/go && sudo tar -C /usr/local -xzf go1.20.8.linux-amd64.tar.gz
+```
+
+Add Go to the environment variable (You may want to change `.zshrc` or `.bashrc` file to permanently add this folder to the `PATH` environment variable)
+```shell
+export PATH=$PATH:/usr/local/go/bin
+```
+
+Verify Go is available
+```shell
+go version
+# go version go1.18.3 linux/amd64
+```
+
+### 2. (Optional) Tune the network
+
+Adjust the TCP window size. If the edge node is far away, we will need to adjust the TCP window size so that the connection can speed up to the speed limit faster. (You could calculate the best TCP window size using https://www.speedguide.net/bdp.php later)
+
+```shell
+cat <<EOT | sudo tee -a /etc/sysctl.conf > /dev/null
+net.core.wmem_max=125829120
+net.core.rmem_max=125829120
+net.ipv4.tcp_rmem= 10240 87380 125829120
+net.ipv4.tcp_wmem= 10240 87380 125829120
+net.ipv4.tcp_window_scaling = 1
+net.ipv4.tcp_timestamps = 1
+net.ipv4.tcp_sack = 1
+net.ipv4.tcp_no_metrics_save = 1
+net.core.netdev_max_backlog = 10000
+EOT
+sudo sysctl -p
+```
+
+### 3. Clone and Build
+Clone the Starlight repository
+```shell
+git clone https://github.com/mc256/starlight.git && \
+cd starlight && \ 
+git checkout v0.3.1
+```
+
+Build the snapshotter plugin and CLI tool
+```shell
+make starlight-daemon ctr-starlight
+```
+
+### 4. Configure Starlight Snapshotter
+
+Find out the IP address / DNS of the Starlight Proxy server and set these two environment variables (Don't Copy-Paste!)
+```shell
+# DO NOT COPY! 
+# This is just an example !!! Get the real address of your server !!!
+echo "<ip address> cloud.cluster.local" | sudo tee -a /etc/hosts > /dev/null
+````
+
+Verify that the Starlight proxy is accessible from the worker. 
+```shell
+curl http://cloud.cluster.local:8090/
+# {"status":"OK","code":200,"message":"Starlight Proxy"}
+curl http://cloud.cluster.local:5000/v2/
+# {}
+```
+If it does not work, please check the firewall configurations, 
+Please add port 8090 and 5000 to the firewall whitelist, the worker has to access these ports.
+
+Install Starlight Snapshotter `systemd` service and CLI tool.
+```shell
+sudo make install install-systemd-service
+```
+
+Enable Starlight snapshotter service
+```shell
+sudo systemctl enable starlight   && \
+sudo systemctl start starlight
+```
+
+Verify Starlight is running
+```shell
+sudo systemctl status starlight
+# it should be "active".
+```
+
+Add Starlight Proxy profile to the Snapshotter's configuration file
+```shell
+sudo ctr-starlight add myproxy http cloud.cluster.local:8090
+```
+
+Confirm that the proxy has been added
+```shell
+sudo ctr-starlight ls
+# [starlight-shared] https://starlight.yuri.moe
+#          [myproxy] http://cloud.cluster.local:8090
+```
+
+Test the proxy is working
+```shell
+sudo ctr-starlight test myproxy
+# ping test success: ok! - http://cloud.cluster.local:8090
+# latency: XX ms
+```
+
+
+### 5. Configure `contaienrd`
+
+Add configuration to `/etc/containerd/config.toml`. 
+(If you have set other `proxy_plugins`, please manually edit the file)
+```shell
+sudo mkdir /etc/containerd/ && \
+cat <<EOT | sudo tee -a /etc/containerd/config.toml > /dev/null
+  [proxy_plugins]
+    [proxy_plugins.starlight]
+      type = "snapshot"
+      address = "/run/starlight/starlight-snapshotter.sock"
+EOT
+```
+
+Restart `containerd` service
+```shell
+sudo systemctl restart containerd
+```
+
+Verify the Starlight snapshotter plugin is functioning
+```shell
+sudo ctr plugin ls | grep starlight 
+# io.containerd.snapshotter.v1    starlight                -              ok
+```
+
+### 6. Convert Container Image
+
+Convert the container image to the **Starlight format** container image and report to the Starlight proxy.
+
+
+```shell
+sudo ctr-starlight convert \
+    --insecure-destination \
+    --notify --profile myproxy \
+    --platform linux/amd64 \
+    docker.io/library/redis:6.2.1 cloud.cluster.local/redis:6.2.1-starlight && \
+sudo ctr-starlight convert \
+    --insecure-destination \
+    --notify --profile myproxy \
+    --platform linux/amd64 \
+    docker.io/library/redis:6.2.2 cloud.cluster.local/redis:6.2.2-starlight
+```
+
+In this example, we load two versions of the Redis container image from docker hub and convert them to the Starlight 
+format container image and notify the Starlight proxy (using `--notify` flag).
+
+### 7. Optimize Container Image
+
+Set `starlight` as the default containerd snapshotter in command line (optional).
+```shell
+export CONTAINERD_SNAPSHOTTER=starlight
+```
+
+Collect traces on the worker for container startup.
+```shell
+sudo ctr-starlight optimizer on && \
+sudo ctr-starlight pull --profile myproxy cloud.cluster.local/redis:6.2.1-starlight && \
+mkdir /tmp/test-redis-data && \
+sudo ctr c create \
+   --snapshotter=starlight \
+   --mount type=bind,src=/tmp/test-redis-data,dst=/data,options=rbind:rw \
+   --env-file ./demo/config/all.env \
+   --net-host \
+   cloud.cluster.local/redis:6.2.1-starlight \
+   instance1 /usr/local/bin/redis-server && \
+sudo ctr task start instance1
+```
+
+You may terminate the container using `Ctrl-C`, and remove the container:
+```shell
+sudo ctr container rm instance1
+```
+
+Repeat the same thing for `redis:6.2.2`
+```shell
+sudo ctr-starlight pull --profile myproxy cloud.cluster.local/redis:6.2.2-starlight && \
+sudo ctr c create \
+   --snapshotter=starlight \
+   --mount type=bind,src=/tmp/test-redis-data,dst=/data,options=rbind:rw \
+   --env-file ./demo/config/all.env \
+   --net-host \
+   cloud.cluster.local/redis:6.2.2-starlight \
+   instance2 /usr/local/bin/redis-server && \
+sudo ctr task start instance2
+```
+
+Terminate the container using `Ctrl-C`, and remove the container:
+```shell
+sudo ctr container rm instance2
+```
+
+Report traces to the Starlight Proxy.
+```shell
+sudo ctr-starlight optimizer off && \
+sudo ctr-starlight report --profile myproxy
+# set optimizer: completed request
+#	sha256:291220ae234f1aa9655359d7e553b05fa9e288fd811b7429f229e5aecd64a181: collected 40.966s file access traces - okay
+#	sha256:561c5b8bb95e26feb56b2bfda1d1fe2aee3229e5d9a2cc879e7524f05ad427a8: collected 13.459s file access traces - okay
+# reported traces: uploaded traces
+```
+
+
+### 8. Clear all the cache and reset the environment
+```shell
+# This script will kill containerd and starlight.
+# And then it removes ALL contents
+# in /var/lib/containerd, /var/lib/starlight, and /tmp/test-redis-data (the redis data directory)
+# and unmount all the mount points in /var/lib/starlight/mounts
+sudo ./demo/reset.sh 
+
+# Confirm that the cache is cleared
+sudo ls -al /var/lib/containerd
+# > ls: cannot access '/var/lib/containerd': No such file or directory
+sudo ls -al /var/lib/starlight
+# > ls: cannot access '/var/lib/starlight': No such file or directory
+
+
+# restart the processes
+sudo systemctl start starlight containerd
+```
+
+### 9. (Optional) Check the metadata database
+
+You could also inspect the metadata database using the Adminer Web UI at `http://<public ip address of the cloud>:8080/` (system: `PostgreQL`,server: `db`, username: `postgres`, password is the same as the username).
+
+In the `file` table, you could see it records the access order of the file.
+
+![adminer](./images/metadatadb-screenshot.png)
+
+
+### 10. Deploying and update container
+
+Start a container using Starlight (it should be much faster)
+```shell
+sudo ctr-starlight pull --profile myproxy cloud.cluster.local/redis:6.2.1-starlight && \
+mkdir /tmp/test-redis-data && \
+sudo ctr c create \
+   --snapshotter=starlight \
+   --mount type=bind,src=/tmp/test-redis-data,dst=/data,options=rbind:rw \
+   --env-file ./demo/config/all.env \
+   --net-host \
+   cloud.cluster.local/redis:6.2.1-starlight \
+   instance3 /usr/local/bin/redis-server && \
+sudo ctr task start instance3
+```
+
+Update a container using Starlight (also way faster)
+```shell
+sudo ctr-starlight pull --profile myproxy cloud.cluster.local/redis:6.2.2-starlight && \
+sudo ctr c create \
+   --snapshotter=starlight \
+   --mount type=bind,src=/tmp/test-redis-data,dst=/data,options=rbind:rw \
+   --env-file ./demo/config/all.env \
+   --net-host \
+   cloud.cluster.local/redis:6.2.2-starlight \
+   instance4 /usr/local/bin/redis-server && \
+sudo ctr task start instance4
+```