The world refusing to start with the standard Data set (no CoA DBC) is now
reported as GameDataMismatch. A Settings card shows the folder in use and
changes it while the server is stopped, using the install checks.
Second part of the audit of every screen for a Docker server. What it found, and
what now happens:
- Play with friends. "Servers open" read the BindIP of the configuration files, which
the containers force to 0.0.0.0, so it was green on a fresh server while who can
connect is decided by the address the ports are published on (bindAddress in
docker.json). Opening and closing now edit bindAddress (and set
CoA.AllowRemoteClients), the open state is read from it, and enabling a mode no
longer fails on netsh / PowerShell after it already rewrote the files. There is no
firewall row on Linux (Docker publishes the ports itself); the Windows reply is
unchanged.
- The "private services" check (Friends and Diagnostics) could never fail on Linux: it
reads the Windows listener table. For a Docker server it is computed from how the
ports are published: the database never, the console only on 127.0.0.1, the game
servers on bindAddress.
- Console, Database tab: read a mysql log file that does not exist; it shows
`docker logs` of the database container.
- The Overview showed a service that had stopped by itself as "Stopped"; it now says
"Stopped unexpectedly".
- Settings: a Docker server that was added but never started had no
worldserver.conf and the page failed; it is created from its template. A talent
file path outside the Core folder is refused (the container cannot see it).
- Backups of configuration now include docker.json; verifying the files of an
imported Docker server looked for worldserver.exe and reported it missing.
- Linux locations: the friend package and the diagnostics go to the desktop folder
(from the user directories file) instead of /tmp, the problem report names the
distribution, and the size advice for companions reads /proc/meminfo instead of
assuming 8 GB.
Every Windows branch is the code that was there; the dispatch is the early return on
docker::is_docker (or the host flavor for the firewall).
Found by auditing every screen for a Docker server.
- Server settings: Ra.Enable, Ra.IP and Updates.EnableDatabases were on the page and
written to worldserver.conf, but the containers get AC_RA_ENABLE, AC_RA_IP and
AC_UPDATES_ENABLE_DATABASES, and the core reads those before the file: the change
was accepted and had no effect, and the page showed the file's value while the
container ran another. A Docker server no longer offers them (nor "restore
recommended defaults" for them), and a change forced through is refused with the
reason. The list is docker::FORCED_KEYS, and a test compares it to the variables
the containers really get, so adding one without the other fails.
- `server restart` (and idlerestart) make the core exit with code 2 and expect a
supervisor to start it again; the Docker world had none and showed "crashed". It now
gets `--restart on-failure:3`, applied once the server is up (so a start that fails
stays failed instead of looping); a stop or `server shutdown` (code 0) is never
restarted.
The signed base package holds the MySQL data directory of a Windows repack. On a
Linux computer extract-schema-base now turns it into a Docker server folder whose
database container opens that directory (mysqlData in docker.json), so the release
SQL, the schema contract and export-baseline run with the Docker installation code.
Windows is unchanged: it still bootstraps the repack's own database program.
docker::install is the Linux counterpart of install::install_base and shares none
of its code (the Windows installer and its folder checks are untouched). Same
order of events: signed manifest, folder / space / game data checks, verified
download, extraction into <dest>.installing, database, one rename. Docker must be
usable; a failure removes the containers, the volume and the staging folder.
The databases start from the package's own dumps, Database/baseline/<kind>.sql.zst,
in the format of the Manager's backups, taken in a state where every migration of
the package is applied. The installer imports them, records the migrations as
applied (baseline) in a few bulk statements, and creates the acore account, the
realm port and the console account. Rebuilding from the repository's SQL does not
work: pending migrations such as rev_20260903_01_live_class_baseline are guards
generated from the maintainers' live database and refuse to run on a database
built from the base scripts, whatever the order.
- coa-release export-baseline --server DIR --out DIR produces the dumps from a
prepared server (repack or Docker).
- The game data stays where the person has it (dataDir in docker.json, mounted
read-only), checked at install time.
- ensure_main_configs creates worldserver.conf / authserver.conf from their
.dist templates when missing, at install and at every start, never overwriting.
- Fix (affects the backend merged in 0.6.0): the database container's health check
is now a ping over TCP. On a new volume the image first runs a temporary server
that answers on its socket but takes no network connection, so the container was
healthy about four seconds (10 s against 14 s measured) before TCP worked.
- docs/ARCHITECTURE.md: D13 (installation, package contract) and D14 (how Windows
and Linux code are kept apart).
Tested: 12 new tests (folder rules, settings, package checks, bulk ledger SQL,
failure cleanup with an in-memory docker), and real runs of the whole chain
through examples/docker_install: baseline dumps exported from a prepared
server, package signed with a throwaway key, installed in 171 s, started by the
Manager (database healthy, world ready in 8 s, auth up), console reached with the
account the installer created, 1630 migrations recorded as applied.
Db runs mysql, mysqldump and mysqladmin inside the database container with
docker exec -i when the server folder is a Docker installation, connecting over
TCP on the container's loopback like the repack's commands do on the host. The
password still travels only in MYSQL_PWD. Every caller keeps using Db unchanged
(accounts, backups, population, migrations, the RA service account), and the
host needs no MySQL client. The repack path is untouched: one line before its
tools are looked up. A stopped container or a missing one is reported as a
database that is not running.
Tested: 4 new tests (commands built for the container, no password on a command
line, repack unchanged, error classification), and a real run against the
Docker folder: SQL queries, a quick backup (156 characters tables), restore with
the previous database kept, a damaged backup refused, the RA service account
created by SQL, the world server console reached through its published
loopback port, an account created through it and listed by the Manager.
A server folder with Settings/docker.json is a Docker installation: MySQL, the
auth server and the world server run in containers controlled through the
docker command line. Every other folder is a repack and goes through the
unchanged Windows code. driver::run, process::observe and layout::scan check
for the marker first and hand over to coa_core::docker.
- start: network, runtime image (libraries only, binaries are mounted from
Core/), database healthy, world listening, then auth. A container that stops
while starting reports the cause found in its log (DatabaseNotRunning,
PortInUse, ...). New error code DockerUnavailable (docker missing, stopped,
or not usable by this user).
- stop: world (180 s to save), auth, database; exit code 0 observed.
- observe: container state, health and port readiness mapped to the existing
ServiceState. Readiness is a connection that stays open, because Docker's
proxy accepts connections before the server inside listens.
- passwords go by environment, never on a command line; the remote console is
published on 127.0.0.1 only; the server's own database updater is turned off
(it needs a mysql client the image does not have and stops the server).
Tested: 18 new tests with an in-memory Docker, and a real run through
examples/drive: database, world (ready in 11 s) and auth started from a
package-shaped folder, stopped with exit code 0 and restarted.
Not included: installation, database setup, backups and SQL through the
container, updates, firewall/exposure, the client under Wine/Proton.