Support for timeout, perCallTimeout for Retry Mechanism about node-sdk-core HOT 4 CLOSED

Hi @kvv002

We do support a general timeout parameter in the constructor of the SDK, which should be sufficient for your use case, as the timeout should be applied to each retried request. Is there a reason that would not be sufficient?

from node-sdk-core.

Hi @dpopp07 , we have the below requirements:

1. maxRetry
2. retryDelay
3. status codes to retry, including E-errors
4. timeout for each retry
5. total timeout including all the retries

With the SDK we were able to solve maxRetry and retryDelay, but for the 3rd point we need to have a configurable retries for E-errors for eg: 5 retries for these errors : [429, 502, 503, 504, 'ECONNRESET', 'ETIMEDOUT', 'EHOSTUNREACH', 'EAI_AGAIN', 'ECONNREFUSED', 'EAI_AGAIN', 'EPIPE']
We weren't able to solve the 3rd point with SDK alone, hence we switched to retry-axios library, using the below config with retry-axios we were able to solve point 1,2,3

const maxRetry = 5
const { IamAuthenticator } = Authenticator
const authenticator = new IamAuthenticator({
    apikey: CLOUDANT_APIKEY,
    timeout: 5000 //does not consider axios (5) retries for E-errors
  })

const service = new CloudantV1({
authenticator: authenticator
})
service.setServiceUrl(CLOUDANT_URL);

service.requestWrapperInstance.axiosInstance.defaults.raxConfig = {
  // ... retry-axios config options
  retry: maxRetry, 
  backoffType: 'static',
  retryDelay: 5000, 
  timeout: 5000, //does not time out
  statusCodesToRetry: [[100, 199], [400, 429], [500, 599]],
  noResponseRetries: maxRetry,
  onRetryAttempt: err => {
    const cfg = rax.getConfig(err);
    console.log('---err--2--', err.code, err.status)
    console.log(`Retry attempt #${cfg.currentRetryAttempt}`);
    const TIME_ELAPSED = (new Date() - start) / 1000
    console.log('--time----', TIME_ELAPSED)
  },
  instance: service.requestWrapperInstance.axiosInstance,
};
rax.attach(service.requestWrapperInstance.axiosInstance);

Now when we include the timeout in retry-axios it does not consider that, and if we include noResponseRetries in the SDK for E-errors it does not consider the same.

We need a workaround which can support all the above 6 points

from node-sdk-core.

@kvv002 Since we are using retry-axios to handle retries within the core, we are limited by what they support in their package.

For something like total timeout including all the retries, I don't believe that package supports that. You would need to request that feature be added to that library for there to be support here.

For configuring a per-request timeout, you need to provide a general timeout to the service client constructor:

const service = new CloudantV1({
  authenticator: authenticator,
  timeout: 5000,
})

from node-sdk-core.

I think this issue has been addressed, so closing it out.

from node-sdk-core.