When there is an error during a reflection refresh, such as the cluster went down or it lost connection to the data source, it never retries on schedule. It permanently fails.
We used >20 different data sources in Dremio. It’s difficult and time consuming to go through every failed reflection and manually disable->enable->refresh each one.
It’s also time consuming to monitor reflections daily.
Is there a way to configure Dremio to so that it continuously retries reflection refreshes on the pre-defined schedule?
By default we try 3 times and then give up. This is only if the coordinator is up and the executors are down or source is down. Probably use the “Never Refresh” checkbox and use REST API to refresh when you know the source is up?
Is there a way to change the default retry from 3 times to unlimited retries?
We have undocumented parameters that can change with every version so difficult to manage for you, as I said the best two options I see are
#1 Refresh less frequently as you said the source go offline and you do not use up your 3 tries too soon
#2 Disable automatic refresh and you take control via REST API calls
We use a handful of the undocumented paramters in our cluster, i.e. increasing field size.
What would the parameter name be for this one?
#1 Does not help us as the data should needs to be updated every 2 hours. If the cluster goes down for 6 hours over the weekend I have to manually refresh every reflection in our cluster.
#2 Updating REST API calls is a significant amount of effort and continuous upkeep from engineering.
Here is the parameter to change the value, as I said this can change without notice
What’s the parameter name? It looks like it is missing in your reply.