missions/tests/Azaion.Missions.E2E.Tests/Tests/Resilience/MigratorRestartTests.cs

using System.Diagnostics;
using System.Net;
using Azaion.Missions.E2E.Fixtures;
using Azaion.Missions.E2E.Helpers;
using Npgsql;
using Xunit;

namespace Azaion.Missions.E2E.Tests.Resilience;

/// <summary>
/// NFT-RES-03 and NFT-RES-04 — migrator behaviour across container restarts.
/// Both scenarios drive the SUT via docker compose and rely on the
/// <see cref="ComposeRestartFixture"/> harness; they share one xUnit
/// collection so a failed teardown of NFT-RES-03 does not leak state into
/// NFT-RES-04.
/// Traces: AC-6.4, AC-6.5, AC-6.6, AC-10.5.
/// </summary>
[Collection("MigratorRestart")]
[Trait("Category", "Res")]
[Trait("db_access", "seed-or-assert-only")]
public sealed class MigratorRestartTests : TestBase, IClassFixture<ComposeRestartFixture>
{
    private readonly ComposeRestartFixture _restart;

    public MigratorRestartTests(ComposeRestartFixture restart) => _restart = restart;

    [SkippableFact]
    [Trait("Traces", "AC-6.6,AC-6.4")]
    [Trait("max_ms", "60000")]
    public async Task NFT_RES_03_migrator_is_idempotent_on_container_restart()
    {
        Skip.IfNot(_restart.Enabled,
            "ComposeRestartFixture disabled (COMPOSE_RESTART_ENABLED!=1). " +
            "NFT-RES-03 needs `docker compose restart` access.");

        // Arrange — clean DB so the migrator is not racing with stale data.
        DbResetFixture.ResetDatabase(TestEnvironment.DbSideChannel);
        var schemaBefore = SnapshotPublicSchema();

        // Capture the wall-clock just before the restart so the log slice
        // does not include pre-existing warnings from the first start.
        var restartUtc = DateTime.UtcNow;

        // Act
        Compose("restart missions");
        await WaitForHealthyAsync(TimeSpan.FromSeconds(30));

        // Assert — no NEW errors AT ALL in the restart slice.
        var logs = DockerLogs.Read("missions-sut", restartUtc);
        AssertNoNewErrorLines(logs);

        var schemaAfter = SnapshotPublicSchema();
        Assert.Equal(schemaBefore, schemaAfter);
    }

    [SkippableFact]
    [Trait("Traces", "AC-6.5,AC-10.5")]
    [Trait("max_ms", "120000")]
    public async Task NFT_RES_04_legacy_gps_tables_dropped_on_first_start_and_subsequent_restart_is_noop()
    {
        Skip.IfNot(_restart.Enabled,
            "ComposeRestartFixture disabled (COMPOSE_RESTART_ENABLED!=1). " +
            "NFT-RES-04 needs `docker compose stop|start|restart` access.");

        // Build-time gate — the migrator must contain the post-B9 DROP block.
        // We probe empirically: seed the legacy tables, restart missions,
        // verify they are gone. If they survive, the build pre-dates B9 and
        // we skip with a clear reason.

        // Arrange — stop missions, seed the legacy tables.
        Compose("stop missions");
        ResetAllAndSeedLegacyTables();
        var legacyPresent = LegacyTablesExist();
        Assert.True(legacyPresent, "seed_legacy_gps_tables did not actually create the legacy tables");

        // Act 1 — first start should drop the legacy tables.
        Compose("up -d missions");
        await WaitForHealthyAsync(TimeSpan.FromSeconds(45));

        var legacyAfterFirstStart = LegacyTablesExist();
        Skip.If(legacyAfterFirstStart,
            "Legacy orthophotos/gps_corrections tables still present after first start; " +
            "this build appears to pre-date B9. NFT-RES-04 is a no-op on pre-B9 builds.");

        // Act 2 — restart should be a no-op (no 'does not exist' errors).
        var restartUtc = DateTime.UtcNow;
        Compose("restart missions");
        await WaitForHealthyAsync(TimeSpan.FromSeconds(30));

        // Assert
        Assert.False(LegacyTablesExist(), "legacy tables reappeared after restart");
        var logs = DockerLogs.Read("missions-sut", restartUtc);
        Assert.DoesNotContain("does not exist", logs, StringComparison.OrdinalIgnoreCase);
    }

    private static void ResetAllAndSeedLegacyTables()
    {
        using var conn = new NpgsqlConnection(TestEnvironment.DbSideChannel);
        conn.Open();
        using var cmd = conn.CreateCommand();
        cmd.CommandText = """
            DROP TABLE IF EXISTS orthophotos;
            DROP TABLE IF EXISTS gps_corrections;
            CREATE TABLE orthophotos (
                id UUID PRIMARY KEY,
                payload TEXT NOT NULL DEFAULT ''
            );
            CREATE TABLE gps_corrections (
                id UUID PRIMARY KEY,
                payload TEXT NOT NULL DEFAULT ''
            );
            """;
        cmd.ExecuteNonQuery();
    }

    private static bool LegacyTablesExist()
    {
        using var conn = new NpgsqlConnection(TestEnvironment.DbSideChannel);
        conn.Open();
        using var cmd = conn.CreateCommand();
        cmd.CommandText = """
            SELECT to_regclass('orthophotos')::TEXT, to_regclass('gps_corrections')::TEXT;
            """;
        using var reader = cmd.ExecuteReader();
        reader.Read();
        var ortho = reader.IsDBNull(0) ? null : reader.GetString(0);
        var gpsCorr = reader.IsDBNull(1) ? null : reader.GetString(1);
        return ortho is not null || gpsCorr is not null;
    }

    private static Dictionary<string, string> SnapshotPublicSchema()
    {
        var rows = new Dictionary<string, string>(StringComparer.Ordinal);
        using var conn = new NpgsqlConnection(TestEnvironment.DbSideChannel);
        conn.Open();
        using var cmd = conn.CreateCommand();
        cmd.CommandText = """
            SELECT table_name || '.' || column_name AS key,
                   data_type
            FROM information_schema.columns
            WHERE table_schema = 'public'
            ORDER BY table_name, column_name;
            """;
        using var reader = cmd.ExecuteReader();
        while (reader.Read())
            rows[reader.GetString(0)] = reader.GetString(1);
        return rows;
    }

    private static void AssertNoNewErrorLines(string logs)
    {
        // Each line is independently checked — a stack-trace dump
        // contains exception keywords; an actual ERROR log line does too.
        var bad = logs.Split('\n')
            .Where(line =>
                line.Contains("error", StringComparison.OrdinalIgnoreCase)
                || line.Contains("exception", StringComparison.OrdinalIgnoreCase))
            .ToArray();
        Assert.True(bad.Length == 0,
            $"expected NO new error/exception lines in restart slice; saw {bad.Length}:\n{string.Join("\n", bad)}");
    }

    private async Task WaitForHealthyAsync(TimeSpan timeout)
    {
        using var http = new HttpClient { Timeout = TimeSpan.FromSeconds(2) };
        var deadline = DateTime.UtcNow + timeout;
        while (DateTime.UtcNow < deadline)
        {
            try
            {
                using var resp = await http.GetAsync(new Uri(TestEnvironment.MissionsBaseUrl + "/health"));
                if (resp.StatusCode == HttpStatusCode.OK) return;
            }
            catch (HttpRequestException) { /* not yet listening */ }
            catch (TaskCanceledException) { /* slow first request */ }
            await Task.Delay(500);
        }
        throw new TimeoutException(
            $"missions did not become healthy within {timeout.TotalSeconds:F0}s");
    }

    private void Compose(string subcommand)
    {
        var psi = new ProcessStartInfo("docker",
            $"compose -f {_restart.ComposeFile} {subcommand}")
        {
            RedirectStandardOutput = true,
            RedirectStandardError = true,
            UseShellExecute = false
        };
        using var p = Process.Start(psi)
            ?? throw new InvalidOperationException("docker CLI not available");
        var stdout = p.StandardOutput.ReadToEnd();
        var stderr = p.StandardError.ReadToEnd();
        p.WaitForExit();
        if (p.ExitCode != 0)
            throw new InvalidOperationException(
                $"`docker compose {subcommand}` exited {p.ExitCode}:\nstdout: {stdout}\nstderr: {stderr}");
    }
}