[rbe cache]: configure FindMissingBlobs batch size

The OSS RE client used a fixed 100-digest limit for each
FindMissingBlobs request. High-latency backends can spend a long time in
serial round trips when an action has many digests to check.

Expose the limit as buck2_re_client.find_missing_blobs_batch_size while
preserving the 100-digest default. Clamp zero to one so an invalid value
does not create a degenerate batching path.

Original-Commit: 72aaf61
Source-PR: facebook#1330

Author: bvinc

app/buck2_re_configuration/src/lib.rs

@@ -476,6 +476,8 @@ pub struct Buck2OssReConfiguration {
     pub remote_cache_compression_threshold: Option<usize>,
     /// Maximum number of concurrent upload requests for each action.
     pub max_concurrent_uploads_per_action: Option<usize>,
+    /// Maximum number of digests to ask about in a single FindMissingBlobs RPC.
+    pub find_missing_blobs_batch_size: Option<usize>,
     /// Time that digests are assumed to live in CAS after being touched.
     pub cas_ttl_secs: Option<i64>,
     /// Whether to chunk large remote-cache blobs using FastCDC 2020 and SpliceBlob.
@@ -615,6 +617,10 @@ impl Buck2OssReConfiguration {
                 section: BUCK2_RE_CLIENT_CFG_SECTION,
                 property: "max_concurrent_uploads_per_action",
             })?,
+            find_missing_blobs_batch_size: legacy_config.parse(BuckconfigKeyRef {
+                section: BUCK2_RE_CLIENT_CFG_SECTION,
+                property: "find_missing_blobs_batch_size",
+            })?,
             cas_ttl_secs: legacy_config.parse(BuckconfigKeyRef {
                 section: BUCK2_RE_CLIENT_CFG_SECTION,
                 property: "cas_ttl_secs",

remote_execution/oss/re_grpc/src/client.rs

@@ -1094,6 +1094,8 @@ pub struct RERuntimeOpts {
     max_concurrent_uploads_per_action: Option<usize>,
     /// Time that digests are assumed to live in CAS after being touched.
     cas_ttl_secs: i64,
+    /// Maximum number of digests per FindMissingBlobs RPC.
+    find_missing_blobs_batch_size: usize,
     /// Whether to chunk large remote-cache blobs using FastCDC 2020 and SpliceBlob.
     remote_cache_chunking: bool,
     /// Minimum blob size for remote cache compression.
@@ -2321,6 +2323,10 @@ impl REClientBuilder {
                 // NOTE: This is an arbitrary number because RBE does not return information
                 // on the TTL of the remote blob.
                 cas_ttl_secs: opts.cas_ttl_secs.unwrap_or(3 * 60 * 60),
+                find_missing_blobs_batch_size: opts
+                    .find_missing_blobs_batch_size
+                    .unwrap_or(100)
+                    .max(1),
                 remote_cache_chunking: opts.remote_cache_chunking,
                 remote_cache_compression_threshold,
                 retries,
@@ -4647,6 +4653,7 @@ impl REClient {
         let mut remote_results: HashMap<TDigest, DigestRemoteState> = HashMap::new();
         let mut digests_to_check: Vec<TDigest> = Vec::new();
 
+        let batch_size = self.runtime_opts.find_missing_blobs_batch_size;
         let mut digest_iter = request.digests.iter();
         while digest_iter.len() > 0 {
             // Sort our blobs based on what action we need to take
@@ -4664,7 +4671,7 @@ impl REClient {
                         // We can check this blob
                         digests_to_check.push(digest.clone());
                     }
-                    if digests_to_check.len() >= 100 {
+                    if digests_to_check.len() >= batch_size {
                         break;
                     }
                 }
@@ -6916,6 +6923,7 @@ mod tests {
             request_metadata_tool_name: DEFAULT_REQUEST_METADATA_TOOL_NAME.to_owned(),
             max_concurrent_uploads_per_action: None,
             cas_ttl_secs: 0,
+            find_missing_blobs_batch_size: 100,
             remote_cache_chunking: false,
             remote_cache_compression_threshold: DEFAULT_REMOTE_CACHE_COMPRESSION_THRESHOLD,
             retries: 0,