Suddenly increase pace of disk usage in Dgraph


I encountered an issue that Dgraph begins to use disk space with high pace and don’t stop this until Dgraph dead.

In particular, my Dgraph ate up 1TB for 2 days.
A sum of p/, w/ and z/ directories before occurring this issue were a degree of 100GB.

Possible causes for this issue includes data exporting and Badger compaction.
A reason for the farmer, I’m exporting all data using /admin/export endpoint for backup at regular intervals.
It seems that increasing disk usage seems began at just time after this task executed (8/17 03:00 UTC+9).

Can you think of anything that might have caused it?

I used v1.0.6 and launched zero’s and server’s into a single node and didn’t use replication.
Now I’m trying to use v1.0.7 to confirm whether reproducing this issue.


Are you using SSD?


Are you using SSD?

Sure, I’m using AWS EC2 Instance and gp2 in storage.