Meilisearch

Memory leak in v1.20-v1.24 causes unbounded growth and OOM kills

critical
Resource ContentionUpdated Oct 6, 2025(via Exa)
Technologies:
How to detect:

Meilisearch versions 1.20 through 1.24 contain a memory leak in the patched LMDB related to parallel post-processing of facets and prefixes. Memory consumption continuously increases over hours until the system runs out of memory, requiring hard resets. Users report memory growing from initial levels to consuming all available RAM (64GB) within 6-12 hours of indexing operations.

Recommended action:

Upgrade to Meilisearch v1.25 or later, which includes the memory leak fix. For versions 1.20-1.24, use temporary workaround: set MEILI_EXPERIMENTAL_NO_EDITION_2024_FOR_FACET_POST_PROCESSING=true and MEILI_EXPERIMENTAL_NO_EDITION_2024_FOR_PREFIX_POST_PROCESSING=true to disable parallel post-processing. Always set Docker memory limits (--memory and --memory-swap) to prevent host system crashes.