Thanks @cabbage.merchant for the information.
How much data are you loading in terms of # rdf (does 72.1M records = rdfs?) and disk size? Could you share your schema file and the data? I am thinking you may have a lot of indexes.
This error basically means that the bulk loader is unable to lease additional UIDs from the Zero within 1 sec and then retries. Eventually, it will make progress if the Zero is alive.
This error means that the Zero is unable to connect to its leader. Question: did you have more than 1 zero in the cluster?
It should not corrupt data. These may be transient connectivity errors.
Yes. Bulk is going to consume a lot of RAM. And we are looking to improve its memory requirements. See Bulk loader xidmap memory optimization for some work in the area.
Please share your heap profile.
My bad about the block profile. You needed to use the --profile_mode block --block_rate 1 options with bulk.
What was your default ulimit -n. In most cases, it is best to bump this number up to 100,000 or so.