bigrquery::bq_auth() Doesn't work, Warning message: Unable to refresh token: invalid_grant Account has been deleted

[benyamindsmith]

I am trying to use the bigrquery package to read and write data to BigQuery.

When I try to get started with the bq_auth() function I get the following error:

> library(bigrquery)
> bq_auth()
Warning message:
Unable to refresh token: invalid_grant
Account has been deleted

Whats the issue here? Do I need to do anything which I missed in the docs?

[jennybc]

This would be pretty surprising as a very first bq_auth() experience, because ... what token is it trying to refresh? 🤔

Can you give any more details about the context?

[benyamindsmith]

@jennybc just following up on this so that if someone else runs into the same issue they can see it here.

This was the first time I installed bigrquery and I wanted to connect to bigquery in a manner that would be similar to how I usually experienced things with googlesheets4.

As a workaround there are 2 approaches:

Possible Workarounds

1. Making a token with gargle

token<-gargle::credentials_user_oauth2(scopes="https://www.googleapis.com/auth/bigquery")
bigrquery::bq_auth(token = token)  

2. Use service account credentials.

Navigate in the GCP console to "IAM & Admin", add/select a relevant account and permissions. Generate an access key and run in R.

bq_token <- gargle::credentials_service_account(
  path = "./data/service.json",
  scopes = "https://www.googleapis.com/auth/bigquery"
)
bigrquery::bq_auth(token=bq_token)

---

I find 2. a better solution for pipeline development as its more friendly to deal with in a Docker setting.

Feel free to close the issue.

[jennybc]

I still don't understand what token the initial attempt was failing to refresh. But I can live with the mystery. The typical user can allow bq_auth() to "just work", i.e. it eventually calls the gargle functions shown above . But yes, the explicit token creation and use you show works. It just shouldn't be necessary, in general.