Repository navigation
GraphLake 1.1.1 improves query consistency, datasource reads and cluster maintenance.
-
Gremlin keeps paged branches on the same input and can query a published CSR generation. Eligible counting traversals retain multiplicity while combining equivalent traversers.
-
Cypher reuses stable correlated CALL plans and preserves aggregate sorting. Cluster SCD2 queries evaluated against the current clock use the live graph. Historical CSR generations use an explicit snapshot time.
-
PostgreSQL frontier and text-key lookups can use native indexes while keeping exact comparisons. Trino schema reads, SQLite conversion, Delta refresh and S3 connection reuse have also been improved.
-
SQL Server can truncate
datetime2to microseconds at the source withtruncate_datetime2_to_microseconds: true. The default checks precision. -
Execution budgets cover queries, Worker stages and background data operations. Schema checks and exploration retain bounded Coordinator capacity. etcd reads retry transient failures, and CSR cleanup retires pointers for inactive model revisions.
-
The Python bindings support concurrent object-version checks for traversals and graph analytics.
Install and upgrade
Download the Debian 12 archive and its SHA-256 file for ARM64 (aarch64) or AMD64 (x86_64). Back up the full catalog, including .graphlake-local/ when present, and keep the previous package for rollback. Rebuild older business-integer SCD2 CSR generations to record their clock behavior, and rebuild a rolled-back model's CSR if its retired pointer has already been collected.