Aurora 26.2 - build 2

Aurora build 2. The jar is the same server as build 1 — no source changes — so this build is about the thing build 1 was missing: it shipped a jar and nothing else. No config, no start script. This one ships both.

What's in the config bundle

Aurora-26.2-config.zip contains server.properties, start.sh and config/aurora-server.yml.

Five networking options Aurora already has, but ships turned off:

Setting	Why
filter-velocity-packet	Upstream's own note says entity velocity packets "can often consume major amounts of network usage, often being up to 60% on large production servers", and filtering keeps the vanilla visual result. That figure is upstream's, measured on their servers — not ours.
filter-move-packets	Drops move packets that wouldn't change anything on the client.
alternative-player-list-tick	Spreads the playerlist tick across buckets instead of spiking one tick.
purpur-alternative-keepalive	Only times a player out if no keepalive got a response in 30s. One dropped packet stops being a disconnect.
flush-location-while-knockback	From Leaf — syncs position immediately on knockback, so combat rubber-bands less.

View distance 10 to 8 and simulation distance 10 to 6. These scale per-player traffic and per-tick work more directly than anything else. Raise them back if your host has headroom.

network-compression-threshold is deliberately left at its default, because whether to raise or lower it depends on whether you're short on bandwidth or CPU.

start.sh carries Aikar's flags — the same set Atlas ships, for the same reason: GC pauses are the one cause of ping spikes the server fully controls. AlwaysPreTouch is deliberately omitted (it makes every boot slow), matching Atlas. It adds --add-modules=jdk.incubator.vector, which Aurora's own startup check asks for.

What a server can and can't do about ping

Ping is mostly the physical distance to the machine and whatever the player's ISP does. No server jar changes that. What it can control is queueing delay, bytes sent back, and GC pauses. Everything above targets one of those three.

Verified

JDK 25.0.2, Apple Silicon:

• Boots with the bundle: Done (6.037s)!, no config errors
• SIMD goes from available ... but are not configured! to detected as functional
• Jar built from commit 512f04e

No ping, TPS or throughput figures are published, because none were measured. Every setting here is chosen for a documented mechanism, not a benchmarked result.

Note on this tag's history

An earlier build 2 was published and retracted: it claimed 64 optimisation patches that did not exist and shipped a jar identical to build 1. Those patch files have been removed (fea5d335) and this tag now points at the clean tree. Sorry for the noise.