jochen
6f1e2f5a0d
A provider failed every consumer for a day and said so only in its journal (hq issue 179). The provisioner loop now emits provisioner.failing after five minutes without a success — create, check or secret — and repeats it every fifteen; provisioner.recovered on the next success, on withdrawal, and on the first success after a restart, so the controller can name it in status (hq ADR 0224).
98 lines
3.2 KiB
Go
98 lines
3.2 KiB
Go
// postgres-provider: postgres's code, one binary the node's runtime launches and speaks MCP to over
|
|
// stdio through the Go SDK (novox/hq ADR 0188, 0193). It serves postgres's tools and the mesh-store
|
|
// seat's verbs and, beside them, runs long: the provisioner that makes postgres the provider of the
|
|
// mesh `postgres-database` interface, and an audit line for each database it provisions or withdraws.
|
|
//
|
|
// stdout is the MCP channel; everything this module says, it says on stderr.
|
|
package main
|
|
|
|
import (
|
|
"context"
|
|
"encoding/json"
|
|
"fmt"
|
|
"os"
|
|
"time"
|
|
|
|
stdio "git.novox.be/novox/mesh-sdk/go"
|
|
)
|
|
|
|
func say(format string, args ...any) {
|
|
fmt.Fprintf(os.Stderr, "[postgres] "+format+"\n", args...)
|
|
}
|
|
|
|
func main() {
|
|
pg, err := ClientFromEnv(os.Getenv)
|
|
if err != nil {
|
|
// Without the server there is nothing to serve and nothing to provision; said, not fatal,
|
|
// so the runtime does not restart a process that cannot do better.
|
|
say("%v; serving no tools and provisioning nothing", err)
|
|
if err := stdio.Serve("", nil); err != nil {
|
|
say("%v", err)
|
|
os.Exit(1)
|
|
}
|
|
return
|
|
}
|
|
if receives := os.Getenv("MESH_RECEIVES"); receives == "" {
|
|
say("MESH_RECEIVES is not set — the provisioner cannot run without it")
|
|
} else {
|
|
h := &Harness{
|
|
Resource: "postgres-database",
|
|
Receives: receives,
|
|
Adapter: provisioner{pg: pg, announce: announce},
|
|
Log: func(format string, args ...any) { fmt.Fprintf(os.Stderr, format+"\n", args...) },
|
|
// A consumer failed for minutes is said on the bus, where the controller hears it and
|
|
// `status` names it (novox/hq ADR 0224).
|
|
Announce: func(event string, body map[string]any) {
|
|
if err := stdio.Emit(event, body); err != nil {
|
|
say("emit %s failed: %v", event, err)
|
|
}
|
|
},
|
|
Node: os.Getenv("MESH_NODE"),
|
|
}
|
|
go h.Run(context.Background())
|
|
}
|
|
go audit()
|
|
if err := stdio.Serve("", Tools(pg)); err != nil {
|
|
say("%v", err)
|
|
os.Exit(1)
|
|
}
|
|
}
|
|
|
|
// announce emits a lifecycle event without letting a broker hiccup fail the provisioning itself.
|
|
func announce(event string, body map[string]string) {
|
|
if err := stdio.Emit(event, body); err != nil {
|
|
say("emit %s failed: %v", event, err)
|
|
}
|
|
}
|
|
|
|
// audit keeps a line for each database granted and withdrawn — observability the provider is best
|
|
// placed to log. The events are its own, read back from its consumer (`consumes`).
|
|
func audit() {
|
|
subscribe := func() error {
|
|
return stdio.Subscribe("postgres.database.*", func(e stdio.Envelope) error {
|
|
var body struct {
|
|
Consumer string `json:"consumer"`
|
|
Database string `json:"database"`
|
|
}
|
|
_ = json.Unmarshal(e.Body, &body)
|
|
switch e.Key {
|
|
case "postgres.database.provisioned":
|
|
say("database provisioned for %s (db %s)", body.Consumer, body.Database)
|
|
case "postgres.database.deprovisioned":
|
|
say("database withdrawn, kept (db %s)", body.Database)
|
|
}
|
|
return nil
|
|
})
|
|
}
|
|
// Asked again until the runtime takes it: the first ask may come before Serve is running.
|
|
for wait := time.Second; ; wait = min(wait*2, time.Minute) {
|
|
if err := subscribe(); err == nil {
|
|
say("auditing database lifecycle events")
|
|
return
|
|
} else if wait >= 8*time.Second {
|
|
say("subscribing to the lifecycle events failed, will retry: %v", err)
|
|
}
|
|
time.Sleep(wait)
|
|
}
|
|
}
|