snaprecv: actually use DataSize in EstimatedDataSize

It was previously using a mix of SST size and data size. I got worried that
this might actually regress. Namely, if an SST got finished, the data
previously in DataSize would now be reflected in the size of the SST, but
compression could result in an overall lower number. This is not something the
users of this number would have handled well (we would've passed negative
numbers to the pacer, and as far as I can tell there's no check).

Simply remove this kind of problem by returning the cumulative data size.

As a nice side effect, we also get a taste for both the SST and data size
in the datadriven tests.

Author: tbg

pkg/kv/kvserver/kvstorage/snaprecv/multi_sst_writer.go

@@ -185,10 +185,13 @@ func (msstw *MultiSSTWriter) ReadOne(
 	return nil
 }
 
-// EstimatedDataSize returns the approximation of the written bytes to SSTs
-// (including currSST).
+// EstimatedDataSize returns the sum of lengths of keys and values passed
+// to any past or current SST. This is monotonically increasing.
+//
+// Note that this is not the same as the size of the SSTs themselves, as
+// SST files carry additional data but also use compression.
 func (msstw *MultiSSTWriter) EstimatedDataSize() int64 {
-	return msstw.sstSize + msstw.currSST.DataSize
+	return msstw.dataSize + msstw.currSST.DataSize
 }
 
 func (msstw *MultiSSTWriter) emitRangeKey(key rangekey.Span) {

pkg/kv/kvserver/kvstorage/snaprecv/testdata/echotest/TestMultiSSTWriterInitSST_interesting=false

@@ -3,7 +3,7 @@
 echo
 ----
 >> finishing msstw
->> sstSize=4623 estDataSize=4623
+>> sstSize=4623 estDataSize=126
 >> sst0:
 rangedel: /Local/RangeID/100/{r""-s""}
 rangekeydel: /Local/RangeID/100/{r""-s""}

pkg/kv/kvserver/kvstorage/snaprecv/testdata/echotest/TestMultiSSTWriterInitSST_interesting=true

@@ -5,11 +5,11 @@ echo
 >> rangekeyset [/Local/RangeID/100/r/AbortSpan/"00000000-0000-0000-0000-000000000000"-/Local/RangeID/100/r/AbortSpan/"6ba7b810-9dad-11d1-80b4-00c04fd430c8")
 >> sstSize=0 estDataSize=94
 >> rangekeyset [/Local/Range"d"/RangeDescriptor-/Local/Range"f"/RangeDescriptor)
->> sstSize=1128 estDataSize=1160
+>> sstSize=1128 estDataSize=403
 >> rangekeyset ["e"-"f")
->> sstSize=4140 estDataSize=4140
+>> sstSize=4140 estDataSize=581
 >> finishing msstw
->> sstSize=5092 estDataSize=5092
+>> sstSize=5092 estDataSize=585
 >> sst0:
 rangedel: /Local/RangeID/100/r{""-/AbortSpan/"00000000-0000-0000-0000-000000000000"}
 rangedel: /Local/RangeID/100/r/AbortSpan/"{00000000-0000-0000-0000-000000000000"-6ba7b810-9dad-11d1-80b4-00c04fd430c8"}