meilisearch.http.request.duration

HTTP request latency distribution

Dimensions:None

Technical Annotations (25)

Configuration Parameters (7)

max_payload_sizerecommended: 100MB (default, can be increased)

Maximum size for batch document uploads

max_indexing_threadsrecommended: 2

limit concurrent indexing to prevent resource exhaustion on large indexes

max_indexing_memoryrecommended: 4 GiB

cap memory usage during indexing to prevent system-wide resource contention

MAX_MAP_COUNTrecommended: 262144

System-level memory-mapped file limit for handling bulk document operations

NOFILE_SOFTrecommended: 65536

Soft file descriptor limit required for high-volume document indexing

NOFILE_HARDrecommended: 65536

Hard file descriptor limit required for high-volume document indexing

client_request_timeoutrecommended: Duration::ZERO (disable) or >5s

Actix timeout that triggers HTTP 408 when workers are unavailable

Error Signatures (6)

MeiliSearch::TimeoutError: The request was not processed in the expected timeexception

Net::ReadTimeout with #<TCPSocket:(closed)>exception

408http status

Request Timeouthttp status

Idle timeout reached for "http://127.0.0.1:7700/indexes/*/search"error code

Expected response code 200, got 408log pattern

Technical References (12)

meilisearch-railscomponenthttp_request.rbfile pathactix-webcomponentdockercomponentOpenTelemetry specificationprotocolPrometheus histogramcomponentcardinalityconcept/tasks/*componenttokenizercomponentsoft space separatorconcept/searchcomponentactix worker threadscomponent

Related Insights (10)

Meilisearch Cloud timeout errors discard indexing jobswarning

▸

HTTP 408 timeout errors during concurrent indexing and querieswarning

▸

Docker deployment has performance overhead versus standaloneinfo

▸

API response time exceeds 2 secondswarning

▸

Unconfigured indexing resource limits cause periodic service hangs with large indexescritical

▸

Bulk document addition causes complete API freezecritical

▸

HTTP request duration histogram fails to track requests exceeding 1 secondwarning

▸

High-cardinality task metrics overwhelm Prometheuscritical

▸

Hash character (#) in queries causes 100x+ performance degradationcritical

▸

HTTP 408 errors with large indexes under concurrent loadwarning

▸