Is it possible to run dgraph bulk several times concurrently?


I am considering migrating to dgraph and I am running dgraph bulk sequentially with little changes in the schema in order to import the result and run some tests (e.g. seeing if having an attribute with multiple indexes affects performance and adding new indexes as I need them).

Since the bulk process takes some time, I’m wondering if I can save some time by running dgraph bulk concurrently pointing to the same zero instance (and specifying different schemas -s <schema> and --out <directory>). Is it safe to do it or could it lead to any trouble in the generated output?

Thanks in advance!

Dgraph uses the Zero instance to request uid lease. So, for each new bulk you gonna need a new Zero instance.

What you can try to do is change the instance offset.

Thanks @MichelDiz, really useful. That helps me understanding better the dgraph internals.