Metadata queries

Do you get this error in the Profile → Error? I managed to dig up the error below after some digging through profiles and Yarn logs:

dataset listing failed: java.util.concurrent.TimeoutException

For us the root cause was a hard coded RPC timeout, which comes into play when your executor is not on the same server as your coordinator (in our case a Yarn executor cluster). When this happens the dataset listing query executes over RPC on the coordinator; vs when the executor and coordinator are the same box and it executes locally (which has no timeout).

com.dremio.service.listing.DatasetListingInvoker.

final ProtocolBuilder builder = ProtocolBuilder.builder()
    .protocolId(57)
    .allocator(allocator)
    .name("dataset-listing-rpc")
    .timeout(10 * 1000);

The above code sets a hard coded timeout of 10 seconds. We found that with large numbers of VDS, some queries take 4 minutes to complete over RPC, though only a few seconds when the coordinator and executor are the same box. We are working to expose this timeout as a support key in our local code base.