summaryrefslogtreecommitdiff
diff options
context:
space:
mode:
authorPierlauro Sciarelli <pierlauro.sciarelli@mongodb.com>2022-07-27 12:09:08 +0000
committerEvergreen Agent <no-reply@evergreen.mongodb.com>2022-07-27 13:26:01 +0000
commit7438eb5acd7f99ad7e1a618eb13598f3e5e5cff4 (patch)
tree559ddb00061cf39c7b80eb2f3e421e0f56a053bf
parent921c4514352f572bc64b78f068d765206c12e301 (diff)
downloadmongo-7438eb5acd7f99ad7e1a618eb13598f3e5e5cff4.tar.gz
SERVER-68130 adjust autoSplitVector's estimatedAdditionalBytesPerItemInBSONArray
-rw-r--r--src/mongo/db/s/auto_split_vector.cpp11
1 files changed, 10 insertions, 1 deletions
diff --git a/src/mongo/db/s/auto_split_vector.cpp b/src/mongo/db/s/auto_split_vector.cpp
index a430d84593c..adbe4ca5818 100644
--- a/src/mongo/db/s/auto_split_vector.cpp
+++ b/src/mongo/db/s/auto_split_vector.cpp
@@ -49,7 +49,16 @@
namespace mongo {
namespace {
-constexpr int estimatedAdditionalBytesPerItemInBSONArray{2};
+/*
+ * BSON arrays are serialized as BSON objects with the index of each element as a string key: for
+ * example, the array ["a","b","c"] is going to be serialized as {"0":"a","1":"b","2":"c"}. The
+ * minimum size for a BSON object is `BSONObj::kMinBSONLength`.
+ *
+ * Given that the `vector<BSONObj>` returned by `autoSplitVector` can't be greater than 16MB when
+ * serialized, pessimistically assume that each key occupies the highest possible number of bytes.
+ */
+const int estimatedAdditionalBytesPerItemInBSONArray{
+ (int)std::to_string(BSONObjMaxUserSize / BSONObj::kMinBSONLength).length()};
constexpr int kMaxSplitPointsToReposition{3};