NoseyNick's Nebula - an Artemis SBS server-in-the-cloud

"Nebula" is my Artemis SBS server in the cloud. Specifically, it's a bunch of scripts that can kick off an AWS or Google Cloud server running a fresh untouched Ubuntu Linux, will install WINE and a bunch of other pre-requisites, install and configure Artemis, run it, click "server", and be ready for your players to connect to within 5 minutes. You can then connect using SSH in order to configure Siege/Single Front/Double Front/Deep Strike/etc, Difficulty Level, Terrain/LethalTerrain/FriendlyShips/Monsters/Anomalies, and click "Start Game" - there are some commands installed to "click buttons for you" or you can run VNC to see / click the buttons yourself.

Contents...

• Step 1: AWS account (once)
• Step 2: Pick your region and VM size
• Step 3: Start your VM
• Step 4: Wait for it to start
• Step 5: Play!
• Step 6: Shutdown
• How It Works
• Optional Extras
• Performance
• Coming Soon? Maybe?
• (See also my other Artemis tools)

Step 1: AWS account (once)

You'll need to get yourself an AWS account ("create a free account"), unless you already have one of course!

You'll want an SSH keypair in the right region (see below). If you're on Linux you probably already know what SSH is. If you're on Windows you'll probably want PuTTY Secure SHell - this is how you'll connect securely across the internet to your Nebula server

You'll want a Network Security Group in the right region. Call it "2010", add inbound rule, protocol TCP, port 22, source "Anywhere" (unless you want to be really specific about where you will be connecting from for administration?), add inbound rule, protocol TCP, port 2010, source "Anywhere" (unless you know all the IPs of all your players?!?), Done. This will allow players to connect to the Artemis server (which runs on TCP port 2010)

There's no cost for any of the above SO FAR

Step 2: Pick your region and VM size

Unfortunately an Artemis server requires a VM with a spec that's too big for the AWS "free tier". Your VM is going to cost you some money to run, however I've found a spec that seems to run Artemis OK^[*] and supports a reasonable number of clients/players for 17c/hr. I think 50c for 3hrs of Artemis is OK, isn't it? Remember how far 50c used to go at the arcade? When you're done with it you can completely nuke it, there's no disks to keep around so no cost when NOT using the server.

Your "region" should probably be chosen according to where most of your players are located, with a bit of an eye on price. Start at AWS's "on demand" pricing page, 4 vCPUs, usually the "compute optimised" are cheapest. You should obviously check for yourself, I am not going to pay your bill for you, but last time I checked some appropriate prices were:

0.17 USD/hr: vCPU=4 ECU=16 memoryGiB=8 storageGB=ebsonly region=ap-south-1 type=computeCurrentGen size=c5.xlarge
0.17 USD/hr: vCPU=4 ECU=16 memoryGiB=8 storageGB=ebsonly region=us-east-1 type=computeCurrentGen size=c5.xlarge
0.17 USD/hr: vCPU=4 ECU=16 memoryGiB=8 storageGB=ebsonly region=us-east-2 type=computeCurrentGen size=c5.xlarge
0.17 USD/hr: vCPU=4 ECU=16 memoryGiB=8 storageGB=ebsonly region=us-west-2 type=computeCurrentGen size=c5.xlarge
0.186 USD/hr: vCPU=4 ECU=16 memoryGiB=8 storageGB=ebsonly region=ca-central-1 type=computeCurrentGen size=c5.xlarge
0.192 USD/hr: vCPU=4 ECU=16 memoryGiB=8 storageGB=ebsonly region=eu-west-1 type=computeCurrentGen size=c5.xlarge
0.194 USD/hr: vCPU=4 ECU=16 memoryGiB=8 storageGB=ebsonly region=eu-central-1 type=computeCurrentGen size=c5.xlarge
0.196 USD/hr: vCPU=4 ECU=16 memoryGiB=8 storageGB=ebsonly region=ap-southeast-1 type=computeCurrentGen size=c5.xlarge
0.202 USD/hr: vCPU=4 ECU=16 memoryGiB=8 storageGB=ebsonly region=eu-west-2 type=computeCurrentGen size=c5.xlarge
0.202 USD/hr: vCPU=4 ECU=16 memoryGiB=8 storageGB=ebsonly region=eu-west-3 type=computeCurrentGen size=c5.xlarge
0.212 USD/hr: vCPU=4 ECU=16 memoryGiB=8 storageGB=ebsonly region=us-west-1 type=computeCurrentGen size=c5.xlarge
0.222 USD/hr: vCPU=4 ECU=16 memoryGiB=8 storageGB=ebsonly region=ap-southeast-2 type=computeCurrentGen size=c5.xlarge
0.227 USD/hr: vCPU=4 ECU=16 memoryGiB=7.5 storageGB=ebsonly region=ap-northeast-2 type=computeCurrentGen size=c4.xlarge
0.2432 USD/hr: vCPU=4 ECU=variable memoryGiB=16 storageGB=ebsonly region=ap-northeast-1 type=generalCurrentGen size=t2.xlarge OR
0.252 USD/hr: vCPU=4 ECU=16 memoryGiB=7.5 storageGB=ebsonly region=ap-northeast-1 type=computeCurrentGen size=c4.xlarge
0.262 USD/hr: vCPU=4 ECU=16 memoryGiB=8 storageGB=ebsonly region=sa-east-1 type=computeCurrentGen size=c5.xlarge

Step 3: Start your VM

SSHKEY=my-ssh-key ./start.sh 271ben

Step 4: Wait for it to start

You should be able to connect to it (using SSH or PuTTY) within about a minute, for example if your Public IP was 192.0.2.3:

ssh -L5900:127.1:5900 ubuntu@192.0.2.3   # <-- use the real IP address here
tail -1000f /var/log/cloud-init.log      # to watch boot progress

If using PuTTY, you'll want to connect to the supplied IP, user "ubuntu", and arrange for a "local port forward" of port 5900 to 127.0.0.1 5900

If it works, you'll see a lot of headers like:

######################################################################
#### YYYY-MM-DD HH:MM:SS : APT updates and installs...
######################################################################

#### YYYY-MM-DD HH:MM:SS : Waiting 30 secs for server...
#### YYYY-MM-DD HH:MM:SS : Waiting 29 secs for server...
#### YYYY-MM-DD HH:MM:SS : Waiting 28 secs for server...
tcp        0      0 0.0.0.0:2010            0.0.0.0:*               LISTEN
######################################################################
#### YYYY-MM-DD HH:MM:SS : DONE - stop the clock!
######################################################################

wget -T3 -O- http://noseynick.org/artemis/aws/init.sh | sudo NAME=271ben bash

Step 5: Play!

Once you see "tcp... 0.0.0.0:2010" you know that you, and your other players, can point your Artemis clients at the Public IP address.

You're going to need to choose game type, skill level, Terrain/LethalTerrain/FriendlyShips/Monsters/Anomalies etc though, and you're going to need to start the game. There are 2 ways to do this:

1. Point a VNC client at 127.0.0.1, EG vncviewer 127.0.0.1
2. Run a series of commands to click buttons for you, and (without being able to see them) HOPE it has done what you expected

Some of the special commands available to you are:

• click-SERVER - would click the "SERVER" button, but this has already been done for you if you saw "tcp... 0.0.0.0:2010"
• click-game-solo for a solo game. Again, this has already been done for you, but is available in case you wish to change the automation
• click-game-pvp click-game-coop click-game-scripted similarly available in case you wish to change the automation
• ManyManyMany 7 - this will set up for a Level7 DoubleFront with Many Terrain, Many LethalTerrain, Many FriendlyShips, Many Monsters, Many Anomalies - a combination popular on USN evening games. NOTE: This command assumes you are currently at the default 1/Siege/None/None/None/None/None. Running this command multiple times will just click the same buttons multiple times, leading to unpredictable settings.
• repeat - to repeatedly run any of the other commands below (or indeed any command) - see example below
• click-level+ click-level- to change skill level. Remember you already start on Level 1, so to get to Level 7 you'll need to click-level+ SIX times, EG: repeat 6 click-level+
• click-type+ click-type- cycle through:
  • Siege
  • Single Front
  • Double Front
  • Deep Strike
  • Peacetime
  • Border War*
  • Infestation
  • back to Siege again
  * Note that "Border War" is not well supported at present, because it moves most of the other buttons below and the auto-clicker just knows WHERE to click, not WHAT to click:
• click-terrain+ click-terrain- cycle through Terrain:
  • None
  • Few
  • Some
  • Lots
  • back to Many
• click-lethal+ click-lethal- cycle through Lethal Terrain, as above
• click-friendly+ click-friendly- as above for Friendly Ships
• click-monsters+ click-monsters- as above for Monsters
• click-anomalies+ click-anomalies- as above for Anomalies
• click-time+ click-time+10 click-time- click-time-10 to set a fixed game time, EG for a 45-minute time-limited game:
  repeat 4 click-time+10; repeat 5 click-time+
• click-custom to configure for a custom game - the details of which are left as an exercise for the reader
• click-start when you have configured all the above, and when all your players are ready, to begin the game
• click-endsim mid-game to end the simulation. This tries to hit "Escape" and "End Simulation" on the server. The server will hopefully be displaying your Game Results / stats, which might not be visible unless you were using VNC or Artemis-Puppy / logger.pl to collect stats for you.
• click-endgame will quit that "Game Results" screen. You should be back at the "Server Control Screen" ready to change level / type / etc, BUT remember the game will be set up with your previous settings, so you should be modifying your level/type/terrain etc, you are NOT starting from scratch. In particular, running ManyManyMany at this point would scramble your settings in unpredictable ways, but if you were previously already on ManyManyMany 7 you could click-level+ to bump it to level 8 and click-start to begin the next game
• click-quit to click the quit button in the bottom-left. May need repeating 2 or 3 times to quit completely ("Back" then "Exit to windows")
• load will tell you a bit about the load of the machine, how many connections you have, etc

Some other useful scripts:

• ./TSN-Artemis.ca.pl | nc 127.1 2010 will set all the names, ship types, and shuttle/fighter configurations for the TSN Canadian Fleet. See my main Artemis parser.pl documentation / code
• If Artemis crashes (probably not my fault!), make sure it is dead (ps, kill, etc), wait until there are zero connections (load will tel you), then bash artemis.sh will attempt to re-run Artemis.
• If all else fails, sudo /sbin/poweroff and start another one from scratch?

Some regular Linux commands that may be useful:

• tail -1000f /var/log/cloud-init.log - look at the last 1000 lines of the cloud-init logfile, and keep watching it
• sudo /sbin/poweroff - Unix command to completely shutdown / terminate / nuke the VM. Note that the server has been set up to shutdown after 3 hours as a "safety net" using an "at" job:
• atq - list all of the commands that have been configured to run "at" particular times, EG:
  2 Dow Mth DD HH:MM:SS YYYY a ubuntu
• atrm 1 - to remove the "at" job number 1 in the above list

Step 6: Shutdown

From inside the VM: sudo /sbin/poweroff

From the AWS CLI: aws ec2 terminate-instances --region $REGION --instance-ids $ID

From the AWS web console, make sure you're in the right region (if you forgot, then yes, sorry, you're going to have to check every single region). Click on the VM in the list. Click "Actions", "Instance State", "Terminate".

How It Works

• start.sh is mostly a wrapper around a fairly complicated aws ec2 run-instances command to start the VM, with cloud-init --user-data to fetch and run init.sh. As above, you can do this manually in the AWS web console if you prefer (Step 3 above)
• init.sh is just a big wrapper to fetch and download all the others:
• puppy.sh runs as non-root "ubuntu", and can send load / progress reports back to discord via a bot if you have set $TOKEN and $CHAN (see below). If you haven't set either of these, then this is really just creating a ~/bin/load which will tell you a summary of load (cpu, memory, TCP connections to 2010, etc)
• apt.sh runs as root and installs WINE and a bunch of other pre-reqs. For those who know Ubuntu/Linux: This adds https://dl.winehq.org/wine-builds/ubuntu/, runs apt update, and apt install ntp unzip xserver-xorg x11vnc winehq-stable xdotool jq
• x11.sh runs as root and starts the X11 server. This is Linux's graphical environment. This is a bare-bones X11 with no desktop, no menus, no login greeter, just enough for WINE to display stuff and VNC to be able to view it (see below). I find AWS's "c5.xlarge", though not sold as a "GPU VM", seems to have enough OpenGL / GLX capability to run Artemis quite nicely. I'm not 100% sure why X11 needs to run as root, seems to be a permissions issue I haven't been able to resolve.
• artemis-dl.sh runs as non-root "ubuntu", it downloads Artemis and any requested Mods. Currently supported options are:

  NAME=	Version
  260 or 260vanilla	2.6.0 Vanilla
  260ben	2.6.0 with Ben's Mod 4.2.12 "2.6.0 Legacy Edition"
  263	2.6.3 Vanilla
  270 or 270vanilla	2.7.0 Vanilla
  270ben	2.7.0 with Ben's Mod 4.2.12
  271 or 271vanilla	2.7.1 Vanilla
  271ben (default)	2.7.1 with Ben's Mod 4.2.12
  Want others?	Let me know!

  It also downloads empty.jpg and empty.png which are empty / blank images used later, and parser.pl and TSN-Artemis.ca.pl in case you want to configure the Canadian Fleet, or do any of the other crazy stuff that my Artemis Parser can do
• vnc.sh runs as non-root "ubuntu", starts VNC, which is a reasonably good way to see your X server remotely if you need to remote-control it. It only listens on "localhost", it's not accessible remotely except you tunnelled your local 5900 to the remote 127.0.0.1 5900 so you can VNC over your SSH connection
• artemis.sh runs as non-root "ubuntu".
  • It will edit artemis.ini to set up for an auto-start.
  • It will use empty.png and empty.jpg to replace all the skyboxes and the little mini-asteroids that show movement. This looks a lot more boring but speeds up VNC considerably.
  • It runs a dummy wine regedit /? with no display. This just makes WINE stop asking pointless questions about Mono and other stuff Artemis doesn't need
  • It runs Artemis.exe! That's the point, right? :-) This will take a while loading all the 3D models and stuff, so in the meantime...
  • It makes all the above click-* shortcut commands, ManyManyMany, and repeat. These are all wrappers around xdotool to simulate X11 mouse clicks (and occasionally keystrokes)
  • It repeatedly tries to click-SERVER, every second, for up to 1 minute, until the button actually appears in the right place and gets clicked.
  • At this point, the server should be listening on TCP port 2010.
  • It waits 1 second and does a click-game-solo
  • IF you provided a discord TOKEN and CHAN, it will announce that your server is ready for players to join.
  • As a cost safety-net, it will create an at job to shutdown the server after 3 hours (approx 50c). If you don't like this, look at atq and atrm 1 to remove job number 1

Optional Extras

aws-pricing.pl produces the price list above. It is not guaranteed to be up-to-date or correct or anything, I'm not responsible for any errors on your AWS bill or anything.

A number of environment variables can be set in "Step 3 Start your VM" or if you are re-running manually in step 4. The examples show NAME=271ben or SSHKEY=my-ssh-key. If you're using start.sh from your own machine, you can set them in the environment, or in a creds.sh which will be read by start.sh. If you're using the AWS web console, you'll add them in your "User data" between the | and bash

• NAME=271ben - passed by start.sh via init.sh to artemis-dl.sh to select what Artemis version and mods you wish to run - see the table of NAME values supported
• VNCPASS=sig93air - passed by start.sh via init.sh to vnc.sh - the password that will be used by VNC. Note this is not a security risk, as the VNC server is not accessible over the internet, unless you have already securely connected over SSH using your SSH key.
• REGION=us-east-1 - used only by start.sh to specify what region you will use in AWS.
• IMAGE_ID=ami-6dfe5010 - used only by start.sh to specify which Ubuntu image you will be starting. See the instructions in the top of start.sh to find the latest
• SG_NAME=2010 - used only by start.sh to specify the name of the Network Security Group, which should allow in TCP port 22 (for your SSH access) and 2010 (for Artemis clients)
• SSHKEY=my-ssh-key - used only by start.sh - name of an SSH key you already uploaded to AWS
• INSTANCE_TYPE=c5.xlarge - set in, and used only by start.sh to specify what size VM to use. See step 2 and the pricing table above
• TOKEN, CHAN and TESTCHAN - passed by start.sh from creds.sh via init.sh to puppy.sh and discord.sh so your server can tell your players when it is ready for them to connect. TESTCHAN will also get a load of progress info during boot, and a message every 5 mins to remind you it's running and tell you about server load.
• ASROOT and ASUSER - passed by start.sh to init.sh. If supplied, these are a list of extra scripts for init.sh to download and run as root, or as the ubuntu user. If you need help debugging your server, I may ask you to add extra scripts here to let me log into your server and debug, for example, or if you have special requirements, let me know and I may be able to write extra scripts for you
• DISPLAY=:20 set in init.sh, used in x11.sh, vnc.sh, and artemis.sh (by WINE). You're unlikely to need to change this.
• LC_ALL=C set in init.sh, mostly to stop warnings from a bunch of programs that complain about no locale being set. You're unlikely to need to change this.

Performance

Artemis.exe crashes from time to time, probably no more or no less often than any other Artemis server to be honest :-/

"Load average" stats during typical USN games show the machine using about 2.5 - 3.5 of the 4 CPUs allocated to it, with no "steal time", so Artemis seems to not be bottlenecked. AWS still uses "powers of 2" CPUs, so 2 would be too few, 8 would be too many, 4 seems to be the Goldilocks "just right".

After many / most games, I also get my Artemis-Puppy bot to post a quick survey (see puppy-poll.sh), which looks a bit like...

Artemis-Puppy BOT Thanks Crew! Please rate the performance of this server - it's a bit of an experiment and I'd appreciate feedback: A Marvellous B Pretty good C Fair D Not good E Bad F Awful Total votes so far (2019-02-17, fakes removed The bot itself votes for A-F to make the options appear, making it LOOK like every letter has at least 1 vote per survey ): A 6   B 10   C 2   D 1   E 0   F 0

... so it looks like USUALLY the players agree these servers are "Pretty good" :-)

Coming Soon? Maybe?

• Look at other clouds - AWS is obviously one of the most mature, but is definitely not the cheapest cloud out there. For consideration: 1&1 CloudWays DigitalOcean Linode Vultr. Any other reccommendations? Must support hourly or per-minute billing, I do not need these all month
• Have Artemis-Puppy bot, and/or Comms commands, to set level, game type, etc, and to shutdown
• Auto-shutdown when nobody is playing? What if someone does connect a client, but never starts a game? Or you finish a game but remain connected doing nothing for hours? Disconnect? Allow some grace time? Allow people to restart the clock if they're still setting up (Comms message "Yes we are still here"?
• Auto-post your Game Results / stats back to the discord channel using parser.pl?
• Collect a game log and submit it back to my archive?
• Play with "spot instances" to further reduce cost? AWS can rudely and suddenly shut you down if someone out-bids you though

(See also my other Artemis tools)

This stuff is maintained by Nick Waterman - Email Me