Ahh @JimWen interesting I’ll give it a look.
@mrjn @ibrahim I was able to get a good heap snapshot- looks like what my snapshot captured is consistent with JimWen’s.
pprof.dgraph.alloc_objects.alloc_space.inuse_objects.inuse_space.061.pb.gz (43.4 KB)
File: dgraph
Build ID: 8ab9db95e603f2648c2702a2bb6c557d41f8348d
Type: inuse_space
Time: May 21, 2020 at 2:44pm (EDT)
Entering interactive mode (type "help" for commands, "o" for options)
(pprof) top
Showing nodes accounting for 15275.58MB, 98.43% of 15519.04MB total
Dropped 204 nodes (cum <= 77.60MB)
Showing top 10 nodes out of 89
flat flat% sum% cum cum%
8438.07MB 54.37% 54.37% 8438.07MB 54.37% go.etcd.io/etcd/raft/raftpb.(*Entry).Unmarshal
3619.64MB 23.32% 77.70% 4685.36MB 30.19% github.com/dgraph-io/badger/v2/table.OpenTable
639.50MB 4.12% 81.82% 1056.52MB 6.81% github.com/dgraph-io/badger/v2/pb.(*TableIndex).Unmarshal
569.12MB 3.67% 85.48% 569.12MB 3.67% github.com/DataDog/zstd.Decompress
519.74MB 3.35% 88.83% 519.74MB 3.35% github.com/dgraph-io/ristretto/z.(*Bloom).Size
440.71MB 2.84% 91.67% 545.22MB 3.51% github.com/dgraph-io/dgraph/protos/pb.(*Mutations).Unmarshal
417.02MB 2.69% 94.36% 417.02MB 2.69% github.com/dgraph-io/badger/v2/pb.(*BlockOffset).Unmarshal
258.37MB 1.66% 96.03% 258.37MB 1.66% github.com/dgraph-io/ristretto.newCmRow
207MB 1.33% 97.36% 207MB 1.33% github.com/dgraph-io/badger/v2/table.NewTableBuilder
166.41MB 1.07% 98.43% 166.41MB 1.07% github.com/dgraph-io/badger/v2/skl.newArena