[RFC] influxdb: replace log.Fatal() by log.Print() #61
Conversation
kb-light
changed the title
|
I didn't test it yet: Will the client send data, as soon as the database becomes reachable again? |
|
On a Fatal/panic kills yanic, so you got a nice log to find the problem. A Print will keep yanic running. |
|
The goal is to keep yanic running, even if the database is not available, to have at least the latest data available for the map. |
database/influxdb/database.go
Outdated
| @@ -114,7 +114,7 @@ func (conn *Connection) addWorker() { | |||
| // create new batch | |||
| timer.Reset(batchTimeout) | |||
| if bp, err = client.NewBatchPoints(bpConfig); err != nil { | |||
| log.Fatal(err) | |||
| log.Print(err) | |||
There was a problem hiding this comment.
NewBatchPoints only fails if the configured precision can not be passed. In that case it will always fail and should panic:
https://github.com/influxdata/influxdb/blob/3b700863eaa3fae65a995ce65ca12fc85c0a6995/client/v2/client.go#L213
| @@ -134,7 +134,7 @@ func (conn *Connection) addWorker() { | |||
| log.Println("saving", len(bp.Points()), "points") | |||
|
|
|||
| if err = conn.client.Write(bp); err != nil { | |||
| log.Fatal(err) | |||
| log.Print(err) | |||
There was a problem hiding this comment.
That should be fine since the InfluxDB client uses net/http that has timeouts and reconnects.
Fatal() calls os.Exit(1), which causes yanic.service to fail if database is not reachable.
| @@ -149,7 +149,7 @@ func (conn *Connection) addWorker() { | |||
| log.Println("saving", len(bp.Points()), "points") | |||
|
|
|||
| if err = conn.client.Write(bp); err != nil { | |||
| log.Fatal(err) | |||
| log.Print(err) | |||
| } | |||
| writeNow = false | |||
| bp = nil | |||
There was a problem hiding this comment.
@corny: Müsste dieses bp = nil nicht in ein else Zweig? Damit die daten nicht verloren gehen?
There was a problem hiding this comment.
Kann man machen, dann stellt sich aber die Frage, wie lange man Punkte sammeln will.
There was a problem hiding this comment.
Mir geht es erst einmal darum, dass yanic für die Auslieferung der aktuellen Daten weiterläuft und sich wieder mit der influxdb verbindet, wenn diese wieder online ist. Dass die Daten aus der Offline-Zeit nicht in der DB landen, entspricht ja auch der Erwartung.
Generell wäre es natürlich schick, wenn man auch diese Daten sammeln kann und nachträglich in der DB anlegt, aber da sehe ich auch das Problem der Zwischenspeicherung.
|
Passt! 👍 |
Fatal() calls os.Exit(1), which causes yanic.service to fail if database is not reachable.