Altered auto-reconnection attempt logic to comply with part 4 of the OCPP2.0.1 standard #221
Conversation
|
This is not quite ready yet, will reopen when it's more fully tested. |
|
I got toxiproxy working, this code should at least pass the pipeline and worked in our implementation. |
ws/websocket.go
Outdated
| @@ -973,7 +985,8 @@ func (client *Client) cleanup() { | |||
|
|
|||
| func (client *Client) handleReconnection() { | |||
| log.Info("started automatic reconnection handler") | |||
| delay := client.timeoutConfig.ReconnectBackoff | |||
| delay := client.timeoutConfig.RetryBackOffWaitMinimum + time.Duration(rand.Intn(client.timeoutConfig.RetryBackOffRandomRange+1)) | |||
There was a problem hiding this comment.
First of all thanks a lot for the contribution. I agree with you that the incremental backoff with range can be used for v1.6 as well, since nothing really speaks against it.
Regarding the implementation, I see the following issues:
- the back-off random range is being added only once
- the range is being added as nanoseconds, which doesn't really add much randomness -> I presume this should be cast to seconds
Design suggestion
Just to quote the specs:
When the Charging Station is reconnecting, after a connection loss, it will use this variable as the maximum value for the random part of the back-off time. It will add a new random value to every increasing back-off time, including the first connection attempt (with this maximum), for the amount of times it will double the previous back-off time. When the maximum number of increments is reached, the Charging Station will keep connecting with the same back-off time.
My interpretation is that for every iteration of the reconnect loop, the new delay should be:
delay *= 2
delay += time.Duration(rand.Intn(client.timeoutConfig.RetryBackOffRandomRange+1)) * time.SecondsDoes this make sense or did you reach a different interpretation?
|
Hi Lorenzo, I'm happy to contribute, thank you for making such a fantastic open source library. As far as both your comments go, you're totally right, they were both oversights on my part, I'm glad you caught them. Both should be resolved by commit c5ef78b Please let me know if you have any more suggestions or find anything else I missed. |
|
LG2M, merged 👍 |
According to part 4 of the OCPP2.0.1 standard: JSON over WebSockets implementation guide, section 5.3. Reconnecting, there are some configuration variables, specifically:
which dictate auto reconnection behaviour of the charge point. I have done my best to implement these changes and update the corresponding network test accordingly.
While the websocket code is shared between both OCPP versions, since OCPP1.6J (to my understanding) doesn't specify auto reconnection behaviour, it shouldn't be incompatible.