truncate split lists in error logs using PrettySample (#6315)

* truncate split lists in error logs using PrettySample

When many splits fail during a search query, the error logs dump
every failed split (with its full error message) into a single log
line. At scale (hundreds of splits per query), this produces multi-KB
unreadable log lines.

Use PrettySample(5) to show only the first 5 items and a count of
the rest, matching the pattern already used for split_offsets in
leaf search. Also adds num_failed_splits/num_splits fields for easy
filtering without parsing the sample.

Files changed:
- root.rs: failed_splits in leaf search response (debug + error)
- list_terms.rs: failed_splits in list terms response
- fetch_docs.rs: split_ids on fetch docs error

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

* fetch_docs: tag each split future with its split_id in error context

Only one split fails when try_join_all returns an error; logging all
split IDs was misleading. Now each future carries its own split_id via
map_err context, so the error message identifies the exact failing split.

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

* fmt: apply nightly rustfmt to fetch_docs.rs

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

---------

Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

Author: fmassot

quickwit/quickwit-search/src/fetch_docs.rs

@@ -16,7 +16,7 @@ use std::collections::{BTreeMap, HashMap};
 use std::sync::Arc;
 
 use anyhow::{Context, Ok};
-use futures::{StreamExt, TryStreamExt};
+use futures::{StreamExt, TryFutureExt, TryStreamExt};
 use itertools::Itertools;
 use quickwit_doc_mapper::DocMapper;
 use quickwit_proto::search::{
@@ -65,32 +65,27 @@ async fn fetch_docs_to_map(
         let split_and_offset = split_offsets_map
             .get(split_id)
             .ok_or_else(|| anyhow::anyhow!("failed to find offset for split {}", split_id))?;
-        split_fetch_docs_futures.push(fetch_docs_in_split(
-            searcher_context.clone(),
-            global_doc_addrs,
-            index_storage.clone(),
-            split_and_offset,
-            doc_mapper.clone(),
-            snippet_request_opt,
-        ));
+        let split_id = split_id.to_string();
+        split_fetch_docs_futures.push(
+            fetch_docs_in_split(
+                searcher_context.clone(),
+                global_doc_addrs,
+                index_storage.clone(),
+                split_and_offset,
+                doc_mapper.clone(),
+                snippet_request_opt,
+            )
+            .map_err(move |e| e.context(format!("split_id={split_id}"))),
+        );
     }
 
-    let split_fetch_docs: Vec<Vec<(GlobalDocAddress, Document)>> = futures::future::try_join_all(
-        split_fetch_docs_futures,
-    )
-    .await
-    .map_err(|error| {
-        let split_ids = splits
-            .iter()
-            .map(|split| split.split_id.clone())
-            .collect_vec();
-        error!(split_ids = ?split_ids, error = ?error, "error when fetching docs in splits");
-        anyhow::anyhow!(
-            "error when fetching docs for splits {:?}: {:?}",
-            split_ids,
-            error
-        )
-    })?;
+    let split_fetch_docs: Vec<Vec<(GlobalDocAddress, Document)>> =
+        futures::future::try_join_all(split_fetch_docs_futures)
+            .await
+            .map_err(|error| {
+                error!(error = ?error, "error when fetching docs in a split");
+                error
+            })?;
 
     let global_doc_addr_to_doc_json: HashMap<GlobalDocAddress, Document> = split_fetch_docs
         .into_iter()

quickwit/quickwit-search/src/list_terms.rs

@@ -141,7 +141,7 @@ pub async fn root_list_terms(
         .collect();
 
     if !failed_splits.is_empty() {
-        error!(failed_splits = ?failed_splits, "leaf search response contains at least one failed split");
+        error!(num_failed_splits = failed_splits.len(), failed_splits = ?PrettySample::new(&failed_splits, 5), "leaf search response contains failed splits");
         let errors: String = failed_splits
             .iter()
             .map(|splits| splits.to_string())

quickwit/quickwit-search/src/root.rs

@@ -778,10 +778,11 @@ pub(crate) async fn search_partial_hits_phase(
         .map_err(|error: TantivyError| crate::SearchError::Internal(error.to_string()))?;
     debug!(
         num_hits = leaf_search_response.num_hits,
-        failed_splits = ?leaf_search_response.failed_splits,
+        num_failed_splits = leaf_search_response.failed_splits.len(),
+        failed_splits = ?PrettySample::new(&leaf_search_response.failed_splits, 5),
         num_attempted_splits = leaf_search_response.num_attempted_splits,
         has_intermediate_aggregation_result = leaf_search_response.intermediate_aggregation_result.is_some(),
-        "Merged leaf search response."
+        "merged leaf search response"
     );
 
     if let Some(resource_stats) = &leaf_search_response.resource_stats
@@ -800,7 +801,7 @@ pub(crate) async fn search_partial_hits_phase(
     }
 
     if !leaf_search_response.failed_splits.is_empty() {
-        quickwit_common::rate_limited_error!(limit_per_min=6, failed_splits = ?leaf_search_response.failed_splits, "leaf search response contains at least one failed split");
+        quickwit_common::rate_limited_error!(limit_per_min=6, num_failed_splits = leaf_search_response.failed_splits.len(), failed_splits = ?PrettySample::new(&leaf_search_response.failed_splits, 5), "leaf search response contains failed splits");
     }
 
     Ok(leaf_search_response)