| 1 | package rollout |
| 2 | |
| 3 | import ( |
| 4 | "context" |
| 5 | "time" |
| 6 | |
| 7 | "github.com/carlosframework/platform/internal/activator" |
| 8 | "github.com/carlosframework/platform/internal/registry" |
| 9 | "github.com/carlosframework/platform/internal/release" |
| 10 | ) |
| 11 | |
| 12 | // executeWorker runs one resolved attempt on a socketless host (spec §6.5, |
| 13 | // §7.9) through the activator's same-entry replacement. The host's good |
| 14 | // build, by its retained identity, goes with it as the fallback for a failed |
| 15 | // adoption in overlap mode; the activator decides whether it applies. |
| 16 | // |
| 17 | // While a hold-ran marker of the app stands on the host, the replacement |
| 18 | // carries no fallback (Astra R3 N5, spec §6.4): a hold build ran there |
| 19 | // without passing and may have migrated the database, so a failed |
| 20 | // replacement must not start the good build against it. |
| 21 | // |
| 22 | // An abandoned result (a staging refusal, a busy entry, a re-check that |
| 23 | // refused) is deferred by done(), as a refused swap is: while the attempt is |
| 24 | // still the host's newest it waits abandonDeferral, and one superseded by a |
| 25 | // newer desire does not wait at all. |
| 26 | func (s *scheduler) executeWorker(ctx context.Context, rt registry.Route, a activator.Attempt) activator.SwapResult { |
| 27 | m := s.m |
| 28 | var fallback *release.StagedIdentity |
| 29 | acct := release.RouteAccount(rt) |
| 30 | if _, marked := holdRanFor(m.d.Book, acct, rt.App, rt.Host); !marked { |
| 31 | if g := rt.Good(); g.Version != "" && g.Digest != "" && g.Version != a.Build { |
| 32 | fallback = &release.StagedIdentity{Version: g.Version, Digest: g.Digest, Demand: g.Demand, |
| 33 | StagedPath: release.VersionPath(m.d.OptDir, acct, g.Version, rt.App)} |
| 34 | } |
| 35 | } |
| 36 | m.emit(rt, EvSwapping, "info", textSwapping(a.Build, a.Cause)) |
| 37 | return m.d.Act.ReplaceWorker(ctx, a, fallback, s.newest) |
| 38 | } |
| 39 | |
| 40 | // admitOne is admit for one host, for the recycle hook: the host's pending |
| 41 | // attempt, if it is a release-class attempt (an adoption) the queue would |
| 42 | // start right now. It honours everything admit does for that host: its |
| 43 | // waits, its cohort's canary and settle, a halt, and swap_concurrency. A |
| 44 | // restart or config attempt is not taken: the recycle runs on the serving |
| 45 | // build, and that attempt follows in its turn. The caller holds the |
| 46 | // scheduler's lock. |
| 47 | func (q *appQueue) admitOne(host string, now time.Time) (dispatch, bool) { |
| 48 | h := q.hosts[host] |
| 49 | if h == nil || h.running != nil || q.running >= q.concurrency { |
| 50 | return dispatch{}, false |
| 51 | } |
| 52 | d, ok := q.gate(h, now) |
| 53 | if !ok || d.Class != classAdopt { |
| 54 | return dispatch{}, false |
| 55 | } |
| 56 | q.start(h, d) |
| 57 | return d, true |
| 58 | } |
| 59 | |
| 60 | // CoalesceRecycle is plan 4's Service.CoalesceRecycle (spec §7.9). The |
| 61 | // activator calls it under its own lock (activator.Service.mu), so it calls |
| 62 | // nothing in the activator and waits on nothing that could: it takes |
| 63 | // Manager.mu (schedulerOf) and then the scheduler's lock, the documented |
| 64 | // order, and the attempt it admits runs on its own goroutine. True means an |
| 65 | // attempt makes the spawn and the recycle is skipped: one admitted now, or |
| 66 | // one already running for the host (ReplaceWorker respawns that child |
| 67 | // itself, and its observation reads a crash as the build's). Once Close has |
| 68 | // begun it is also true, with nothing admitted. |
| 69 | func (m *Manager) CoalesceRecycle(host string) bool { |
| 70 | s := m.schedulerOf(host) |
| 71 | if s == nil { |
| 72 | return false |
| 73 | } |
| 74 | s.mu.Lock() |
| 75 | if h := s.q.hosts[host]; h != nil && h.running != nil { |
| 76 | s.mu.Unlock() |
| 77 | return true |
| 78 | } |
| 79 | // Counted before it is admitted, as the scheduler loop counts its |
| 80 | // admissions, so Close joins its done() (spec §8.6, Task 11 fix round |
| 81 | // 1). joinMu is a leaf, so it may be taken here under activator.Service.mu |
| 82 | // and s.mu. Once Close has begun nothing is admitted and the recycle is |
| 83 | // skipped: the agent is stopping, and the child goes with it. |
| 84 | if !m.enter(true) { |
| 85 | s.mu.Unlock() |
| 86 | return true |
| 87 | } |
| 88 | d, ok := s.q.admitOne(host, m.d.Now()) |
| 89 | if !ok { |
| 90 | m.leave() |
| 91 | s.mu.Unlock() |
| 92 | return false |
| 93 | } |
| 94 | s.mu.Unlock() |
| 95 | go func() { |
| 96 | defer m.leave() |
| 97 | s.execute(m.ctx, d) |
| 98 | }() |
| 99 | return true |
| 100 | } |
| 101 | |
amadan