Skip to content

Releases: hopskipnfall/EmuLinker-K

0.14.0 Don't crash game on drop, better ping, cache optimized

23 Oct 08:31
@hopskipnfall hopskipnfall
0.14
1213896
Compare
Choose a tag to compare
0.14.0 Don't crash game on drop, better ping, cache optimized Latest
Latest

Major changes in this version:

  1. When a user drops from a game, if two or more players remain the game will not stop #133

This is how things worked from version 0.11.2 and before, but was lost when we moved to running the entire server on a single port. This has not been thoroughly tested and there is a rare edge case where the game may pause for up to one minute before continuing. If you notice strange behavior when a game ends, please let us know.

  1. Ping calculation is more accurate

Until now, ping has been measured at only millisecond-level precision and the average ping was calculated by truncating instead of rounding.

  1. Significant performance boost for the game data cache

The protocol uses a cache to avoid sending multiple packets of identical game data. After some inspection I realized that the specialized data structure being used was almost optimally inefficient 😂 I made some small changes which reduce cache lookup time by about 80% in practice. I haven't run the full suite of performance tests for comparison with previous versions, but it's possible this results in a noticeable improvement.

Other changes:

  • I switched from Dagger to Koin for dependency injection. This removes the need for annotation processing, which speeds up build times and removes nearly all generated sources. This was a big hurdle to making it possible to comfortably edit the server code using VS Code.
  • Some game data processing code that was previously synchronized with locks had the locks removed (presumed unnecessary), which may have a performance benefit.
  • Some changes in how dependencies are defined means the jar is close to half the size.
  • I some changes that may mitigate lag spikes experienced when a user joins the server (please provide feedback on this).

As always if you notice bugs feel free to create a GitHub issue or post in the Kaillera Reborn Discord server.

PRs

Full Changelog: 0.13...0.14

Contributors

hopskipnfall
Assets 3
Loading

0.13.2 Lagstat changes and ESF admin client fix

16 Sep 09:28
@hopskipnfall hopskipnfall
0.13
623cd2a
Compare
Choose a tag to compare
0.13.2 Lagstat changes and ESF admin client fix

0.13.2 update: The latest update in this release includes #130 which fixes a bug causing users to be kicked for inactivity incorrectly. It also allows all users to use /lagstat, not just admins and game owners.

I first want to thank everyone who tried out and gave feedback on ELK 0.12.0. As of now there are ELK servers running in:

🇦🇺 Australia
🇧🇷 Brazil (x2)
🇨🇱 Chile
🇨🇳 China (x2)
🇫🇷 France
🇯🇵 Japan
🇳🇿 New Zealand
🇵🇪 Peru
🇿🇦 South Africa
🇸🇪 Sweden
🇺🇸 USA (x4)
🇻🇳 Vietnam

Many ELK users gave feedback (thank you!) that the new lagstat data isn't always helpful. It turns out the lag statistics weren't being calculated correctly for users on higher than 1 frame delay, or when there were more than 2 players in a single room. I took this opportunity to completely redesign the lagstat feature. I'm working on a long-form document describing specifically how lag is measured, but for now here is quick description of how to use it:

17:56:50 <nue> /lagstat
17:56:50 <Server> Total lag over the last 3m: 1821ms
17:56:50 <Server> Lag caused by players: P1: 10ms, P2: 1801ms

The total lag should be a relatively accurate measurement of the total lag that occurred over the measurement period, and should be a good reference point for how objectively laggy a game is. I would be interested to hear feedback on what ranges feel good, laggy, and unplayably laggy. The lag caused by individual players is a measurement of how much lag each player definitively caused. Note that if multiple players are lagging but one is lagging worse, the current lag measurement strategy may only be able to identify that the player causing the most lag is lagging. If the game feels laggy, use the /lagstat command to identify the player causing the most lag and then ask that player to fake ping to a higher frame delay. If lag continues you should repeat the process.

While you can still use /lagreset to reset lag measurements, note that /lagstat now will only show information for the last 3 minutes.

Other improvements:

  • Anybody (not just the game owner) can use both the /lagstat and /lagreset commands.
  • Support for the ESF Admin Client is fixed.
  • The server version will now be listed in server lists (not just "elk").
  • Dependencies have been updated, most importantly we are now using the Kotlin K2 compiler.
  • Some small optimizations in the GameData and CachedGameData action handlers which may provide better JIT compilation on high volume code paths
  • Extended the amount of time between some periodic logic, most significantly checking in with the central server (every 30m instead of 1m) and checking for user timeouts etc.

Full Changelog: 0.12.0-beta...0.13

Assets 3
Loading

0.12.0 Single Port/Netty

24 May 09:49
@hopskipnfall hopskipnfall
0.12.0-beta
f1431c0
Compare
Choose a tag to compare
0.12.0 Single Port/Netty

This version contains some major architectural changes to the code—most significantly running the server off of a single port instead of allocating a port for each user—which results in higher performance on resource-limited hardware.

If you run into issues with this release feel free to look for help in the Discord:

Discord

Feature changes since previous release:

  • The option server.improvedLagstatEnabled has been removed and improved lagstat is now the only option
  • As the server now runs on only one port, the following settings now do nothing: controllers.v086.portRangeStart, controllers.v086.extraPorts
  • When posting to Twitter, the server will now not attempt to post duplicate tweets (as they will fail anyway)
  • When the server shuts down it will now log users out first, notifying them that the server is turning off
  • You can configure a server chat message displayed to users as they enter or create a room using 
    KailleraServerImpl.JoinGameMessage.1=This is the first message
KailleraServerImpl.JoinGameMessage.2=This shows up second
KailleraServerImpl.JoinGameMessage.3=And so on...

Performance

After rigorous testing under high load with simulated clients on a single-core machine, this release is found to outperform EmuLinkerSF v0.93.2 (which this server is based on), EmuLinker-K v0.11.2 (the previous stable release of this server), as well as Emulinker-X v3.6 (another Emulinker fork emphasizing performance improvements).

image

What's Changed

Full Changelog: 0.11.2-beta...0.12.0-beta

Contributors

hopskipnfall
Assets 3
Loading

0.11.2 Twitter integration

17 Jul 08:10
@hopskipnfall hopskipnfall
0.11.2-beta
0cd7bcb
Compare
Choose a tag to compare
0.11.2 Twitter integration

Adds a feature where the server will post to twitter when someone opens up a new game.

This feature is experimental and may be removed or changed at any time. If you like this feature or have feedback, please let us know in the Discord

image

To set it up:

  1. Create a new Twitter account for your server's bot (you can also share one between multiple servers).

  2. Make a new app in the Twitter developer dashboard using the Free tier.

  3. Configure the following values in config/emulinker.cfg:

twitter.enabled=true

# Delay (in seconds) before sending a tweet.
twitter.broadcastDelaySeconds=20

# Comma-separated list of phrases that, if found in the name
# after a "@", will prevent tweet posting.
twitter.preventBroadcastNameSuffixes=waiting,restart
# If true, will simply delete the tweet when the game starts.
twitter.deletePostOnClose=false

# Fill these in from the Twitter app you created:
twitter.auth.oAuthAccessToken=
twitter.auth.oAuthAccessTokenSecret=
twitter.auth.oAuthConsumerKey=
twitter.auth.oAuthConsumerSecret=
  1. You will also need to configure some messages in config/language.properties:
KailleraServerImpl.TweetPendingAnnouncement=Posting a tweet in {0} seconds. Type \"/stop\" to disable.
KailleraServerImpl.TweetCloseMessage=(opponent found)
KailleraServerImpl.CanceledPendingTweet=Canceled pending tweet.

With these settings, users whose name ends in @waiting (meaning they are waiting for a specific person to join their game) or @restart (meaning they are restarting the game and are waiting for the same person to join) will not have tweets sent. Similarly, users will be notified and given 20 seconds to type /stop to stop the tweet from sending. After the game starts, the account will respond to the original tweet with the text "(opponent found)".

What's Changed

Full Changelog: 0.10.7-beta...0.11.1-beta

Contributors

hopskipnfall
Assets 3
Loading

0.10.7 Public Beta Launch

18 May 09:52
@hopskipnfall hopskipnfall
0.10.7-beta
0e33f67
This commit was created on GitHub.com and signed with GitHub’s verified signature. The key has expired.
GPG key ID: 4AEE18F83AFDEB23
Expired
Verified
Learn about vigilant mode.
Compare
Choose a tag to compare
0.10.7 Public Beta Launch

This is the first stable (beta) release of EmuLinker-K, which is an improved version of EmulinkerSF, rewritten in Kotlin.

This release is intended for server admins who are already familiar with running EmuLinker servers. If you are brand new to hosting a server, you might want to skip this release. I will write more detailed instructions on setting up a server after we get some initial feedback.

Launching the server is almost identical to ESF, and it's compatible with your existing ESF config files. There are two important differences from ESF however:

  1. A minimum Java version of 17 is required.
  2. Your config files (emulinker.cfg and language.properties) need to be saved in UTF-8 format. If the file only contains English words, this probably doesn't matter.

There are a few new features in this build, but with this release I am more interested in getting feedback about core functionality. One difference you may notice is that the /lagstat command is overhauled and more accurately tracks lag caused by players. Currently this doesn't count lag spikes, it counts individual lagged frames (classified into "big" and "small" amounts of lag for the frame).

Feedback

Feel free to file bugs and feature requests on this repository in the Issues tab, or find our channel in the Kaillera Reborn discord:

Discord

What's Changed (all PRs since EmuLinkerSF 0.92.4)

Read more

Contributors

hopskipnfall
Assets 3
Loading

0.9.2-alpha Fix several critical latency bugs

24 Apr 11:29
@hopskipnfall hopskipnfall
0.9.2-alpha
687573c
Compare
Choose a tag to compare
0.9.2-alpha Fix several critical latency bugs Pre-release
Pre-release

This fixes a long-standing bug where lag abruptly occurs after a certain amount of time passes (see the description of #70).

v0.6.3 remains the latest "stable" release, but this is probably comparable. More load testing is required to tell for sure.

Known bugs:

  • #72 At the end of load tests, the server lags significantly when users leave the server.
  • #71 Stale/inactive users are not getting timed out correctly.

What's Changed

Full Changelog: 0.9.1-alpha...0.9.2-alpha

Contributors

hopskipnfall
Assets 3
Loading

0.9.1-alpha

19 Feb 10:38
@hopskipnfall hopskipnfall
0.9.1-alpha
a26d9b0
Compare
Choose a tag to compare
0.9.1-alpha Pre-release
Pre-release

Candidate for comparing against 0.6.3.

What's Changed

  • Rename Request and Notification message names as their toString representations are not clear by @hopskipnfall in #60
  • Set maximum parallelism to 1 for TwitterBroadcaster. by @hopskipnfall in #65
  • Improve logging, mark areas for fixing, update metrics logging. by @hopskipnfall in #66
  • Apply changes from #67 to HEAD, improve cleanup logic for users. by @hopskipnfall in #68

Full Changelog: 0.9.0-alpha...0.9.1-alpha

Contributors

hopskipnfall
Assets 3
Loading

0.6.3 Improve monitoring for comparison against stable version

18 Feb 12:19
@hopskipnfall hopskipnfall
0.6.3
33e5a0e
Compare
Choose a tag to compare
0.6.3 Improve monitoring for comparison against stable version Pre-release
Pre-release

This is a version of 0.6.2, but with the change #67 which adds some monitoring for comparison against coroutine-based versions.

Full Changelog: #67

Assets 3
Loading

0.9.0-alpha Remove user events channel buffer

29 Oct 15:36
@hopskipnfall hopskipnfall
0.9.0-alpha
8558ed4
Compare
Choose a tag to compare
0.9.0-alpha Remove user events channel buffer Pre-release
Pre-release

The main change in this release is #59, where an infinite loop for each user waiting on a Channel is removed and replaced with individual ad-hoc coroutines for each user-directed event.

This is release hopes to address a bug present starting in 0.7.0 (specifically #47) where the server dramatically slows down responding to requests after a few hours and general slowness.

The current mostly stable version of the server is 0.6.2.

What's Changed

Full Changelog: 0.8.0-alpha...0.9.0-alpha

Contributors

hopskipnfall
Assets 4
Loading

0.8.0-alpha Additional cleanup and possible optimizations

06 Oct 12:11
@hopskipnfall hopskipnfall
0.8.0-alpha
25f0a23
Compare
Choose a tag to compare
0.8.0-alpha Additional cleanup and possible optimizations Pre-release
Pre-release

This is essentially no different from 0.7.0-alpha, but notably the number of IO dispatchers is greatly reduced (see ServerMain.kt) which might address the issue of the server slowing down after a few hours.

Again, this is a very much alpha release. Testing is appreciated but please know this and the previous release are very experimental.

What's Changed

Full Changelog: 0.7.0-alpha...0.8.0-alpha

Contributors

hopskipnfall
Assets 3
Loading