Graph Node indexes data from blockchains such as Ethereum and serves it over GraphQL
Repo graphprotocol/graph-node, read from GitHub 16 min ago. graphprotocol has not joined Take My Repo.
$ docker pull graphprotocol/graph-node:v0.45.0
GRAPH_GETH_ETH_CALL_ERRORS renamed to GRAPH_RPC_ETH_CALL_ERRORS. The env var (and internal config field) controlling which eth_call error messages are treated as reverts is renamed for provider-neutral naming. Update it before upgrading, or the setting is silently ignored and defaults apply. (#6487)GRAPH_POSTPONE_ATTRIBUTE_INDEX_CREATION (introduced in v0.44.0 but inert — hardcoded off) now actually defers non-critical attribute indexes until a new deployment nears chain head (GRAPH_POSTPONE_INDEXES_CREATION_THRESHOLD, default 10000 blocks from head). Index creation now runs in a background task instead of blocking block processing, and interrupted CREATE INDEX CONCURRENTLY runs are detected and rebuilt on retry instead of being silently skipped. (#6608)ethereum.decodeParams host function. Decodes calldata/event data whose top-level type is a dynamic tuple (e.g. Gnosis Safe's execTransaction), which ethereum.decode cannot handle since it expects a single ABI value with a leading offset word. Gated behind apiVersion 0.0.10; ethereum.decode itself is unchanged. (#6649)EthereumLogFilter merging per-handler receipt flags with overwrite instead of OR semantics, so a handler declaring receipt: true could silently run with event.receipt == null if another handler on the same filter key declared receipt: false. (#6558)unexpected null in handler errors on every event; Graft::validate now rejects these upfront. (#6610)_logs queries returning an empty result (indistinguishable from a subgraph producing no logs) when no log store is configured; now returns a NotSupported error. (#6637)pool_size = 0; such shards' chains are now skipped instead of erroring. (#6648)optimism-mainnet instead of the optimism chain ID used by Firehose endpoints, so Optimism's extended block data went undetected. (#6662)tsvector) search columns being silently dropped when copying or grafting a subgraph, leaving the destination's fulltext index empty. (#6687)StackUnderflow/StackOverflow/OpcodeNotFound EVM halts being misclassified as non-deterministic — Reth surfaces them via its Debug format (e.g. EVM error: StackUnderflow), which didn't match graph-node's space-separated deterministic-error patterns, so subgraphs indexing via Reth-backed RPC providers stalled and retried instead of recording a deterministic revert. (#6645)graphman dump panicking on tables with more than ~2 GiB in a single string/binary column; batches are now split into byte-bounded slices before conversion to Arrow. (#6646)gnd add producing a subgraph that failed codegen/build because it never wrote new event entities to schema.graphql; unified add and init onto one scaffolding path, which also fixes event-name collisions across data sources, overloaded-event disambiguation, small-integer type mapping, and reserved-word parameter names. (#6660)GRAPH_GETH_ETH_CALL_ERRORS, rename it to GRAPH_RPC_ETH_CALL_ERRORS before upgrading. (#6487)GRAPH_POSTPONE_ATTRIBUTE_INDEX_CREATION now actually takes effect (it was a no-op in v0.44.0). If you already set it expecting no behavior change, be aware new deployments will now defer some attribute indexes until near chain head. (#6608)Thanks to all contributors for this release: @cargopete, @dimitrovmaksim, @fordN, @incrypto32, @lutter, @YaroShkvorets
Full Changelog: v0.44.0...v0.45.0
$ docker pull graphprotocol/graph-node:v0.44.0
EntityCache::load_related returned wrong derived-collection membership on same-block parent reassignment. Membership was decided against intermediate cache layers rather than the entity's final state, so entities reassigned to a new parent within the same block could be left in the old collection, omitted from the new one, or wrongly retained after a cross-handler revert. See Upgrade Notes. (#6548)--elasticsearch-url, --elasticsearch-user, --elasticsearch-password (and the matching ELASTICSEARCH_* env vars) are gone. Configure Elasticsearch under the new [log_store] section in graph-node.toml. See Upgrade Notes. (#6278)_logs field exposes subgraph-emitted logs (mapping log.* calls, runtime, system) with filters for level, timestamp range, and text search, plus pagination and orderDirection. The old Elasticsearch sink for subgraph logs is replaced by a unified LogStore abstraction with three backends configured under a new [log_store] section in graph-node.toml: File (JSON Lines), Loki, or Elasticsearch. Omit [log_store] to disable; logs still go to stdout/stderr. (#6278)trace_filter deserialization failing on providers that omit result.output (notably Sonic), which caused repeated trace ingestion retries. Fixed upstream in alloy 2.0.5 (alloy#3931) and picked up via the bump in this release. (#6576)BigInt values: to_signed_u256 replaced by fallible to_i256, and to_unsigned_u256 now errors on values >= 2^256 instead of panicking. (#6560)graphman chain change-shard failing with a duplicate-key error on revert to the original shard. Backups now use unique names and revert reuses the existing <chain>-old. (#6199)EntityCache::load_related results for same-block parent-reassignment edge cases (#6548). Subgraphs that read derived fields within handlers may diverge in POI from v0.43.0 on blocks exhibiting this pattern; resync from before the affected blocks if affected. The v0.44.0 result is canonical.--elasticsearch-* flags or ELASTICSEARCH_* env vars, migrate to a [log_store] section in graph-node.toml before upgrading. See docs/log-store.md. (#6278)Thanks to all contributors for this release: @erayack, @fordN, @incrypto32, @lutter
Full Changelog: v0.43.0...v0.44.0
$ docker pull graphprotocol/graph-node:v0.43.0
skipDuplicates for immutable entities. A new skipDuplicates parameter on the @entity directive (@entity(immutable: true, skipDuplicates: true)) silently skips duplicate inserts instead of failing the subgraph. This unblocks subgraph composition on Amp-powered subgraphs where SQL queries can produce the same entities across block ranges. (#6458)json_rpc_timeout, request_retries, max_block_range_size, polling_interval, block_batch_size, etc.) can now be set per chain in config.toml instead of relying on global environment variables. Fully backwards compatible — env vars remain the fallback. (#6459)[DEBUG-BUILD] prefix on all log lines. (#6488)call_meta table, significantly faster on large call_cache tables. The old --ttl-max-contracts flag has been replaced by --max-contracts, which computes an effective TTL cutoff instead of limiting per-invocation iteration. Eviction now also returns stats on the number of contracts and entries removed. (#6476) (#6477)eth_getBlockByHash and eth_getBlockByNumber RPC calls by checking the block cache first in block_pointer_from_number and fetch_full_block_with_rpc. (#6491, partially reverted by #6537)ChainStore query methods (ancestor_block_ptr, block_parent_ptr) skip deserializing the full block data JSONB column when only hash/number/parent are needed. (#6456)application_name when PGAPPNAME is not set, making it easier to identify graph-node connections in pg_stat_activity.once and polling filters match the same block — only one trigger type was firing. (#6530)endBlock in two cases: the block-skip optimization bypassing the max_end_block check, and max_end_block being set to None when multiple data sources share the same endBlock. (#6474)isDeprecated: false for __InputValue, which caused some client libraries to fail. (#6475)ipc:// or file:// URLs — the URL was passed directly to the transport instead of extracting the file path. (#6443)graphman config pools not working due to hardcoded pool size override. (#6444)gnd indexer command that delegates to graph-indexer, allowing indexer management (allocations, rules, cost models, status) directly through gnd. (#6492)gnd deploy now prompts for --version-label in interactive mode and requires it in non-interactive mode. (#6532)Thanks to all contributors for this release: @aayushbaluni, @dimitrovmaksim, @incrypto32, @isum, @lutter, @SunTzu
Full Changelog: v0.42.1...v0.43.0
mirror_primary_tables holding locks indefinitely, exhausting the connection pool. Added statement_timeout and lock_timeout to prevent stuck queries from cascading into a full lockup. (#6446)ingestor setting in [chains.*] config sections having no effect — block ingestion ran regardless of its value. (#6447)$ docker pull graphprotocol/graph-node:v0.42.0
no_eip2718 provider feature flag. (#6063) (#6317)docs/amp-powered-subgraphs.md for details. (#6218) (#6293) (#6369)docs/implementation/sql-interface.md. (#6172)diesel_async, eliminating blocking database calls from tokio tasks. This improves throughput and reduces the risk of thread starvation under heavy load. (#6194) (#6185)features in the config file. (#6084)current: include in GraphQL requests. (#6293)first and last aggregation support to String, Bytes, and entity reference types. (#6225)GRAPH_STORE_ACCOUNT_LIKE_SCAN_INTERVAL_HOURS, GRAPH_STORE_ACCOUNT_LIKE_MIN_VERSIONS_COUNT, and GRAPH_STORE_ACCOUNT_LIKE_MAX_UNIQUE_RATIO. (#6209)SubgraphIndexingStatus via the index-node API. (#6201)GRAPH_STORE_SETUP_TIMEOUT.GRAPH_STORE_DISABLE_CALL_CACHE, useful for indexers with locally hosted RPC nodes. (#6214)graph-node-debug image with frame pointers and linux-perf. (#6418)std::sync::RwLock with parking_lot::RwLock across multiple components for better performance under contention.ipfs.map() callbacks. (#6336) (#6416)nocase filter generation for non-String types (e.g., Bytes), which caused SQL errors like operator does not exist: bytea ~~* bytea. (#6351)can_copy_from failures not fully rolling back deployment creation, leaving orphaned records. (#6228)histogram_bounds statistics. (#6275)U256 in the runtime — now returns an error instead. (#6219)graphman copy copying pruned earliest_block_number from source, which could leave the destination deployment in an invalid state. (#6384)0x being incorrectly cached in the call cache, preventing stale empty results from persisting. (#6187)graphman dump and graphman restore (experimental). New commands for exporting and importing subgraph data in Parquet format, enabling backup and migration of deployment data across shards. Supports incremental dumps and progress reporting. See docs/dump.md for details. (#6397)graphman chain call-cache remove now supports --ttl-days to remove stale call cache entries that haven't been accessed within a specified number of days. (#6186)gnd as a drop-in replacement for graph-cli: init, add, codegen, build, publish, deploy, create, remove, auth, clean, and test commands. (#6282)Thanks to all contributors for this release: @DaMandal0rian, @dimitrovmaksim, @fubhy, @hudsonhrh, @incrypto32, @isum, @lutter, @shiyasmohd
Full Changelog: v0.41.2...v0.42.0
no_eip2718 provider feature flag. (#6063) (#6317)docs/amp-powered-subgraphs.md for details. (#6218) (#6293) (#6369)docs/implementation/sql-interface.md. (#6172)diesel_async, eliminating blocking database calls from tokio tasks. This improves throughput and reduces the risk of thread starvation under heavy load. (#6194) (#6185)features in the config file. (#6084)current: include in GraphQL requests. (#6293)first and last aggregation support to String, Bytes, and entity reference types. (#6225)GRAPH_STORE_ACCOUNT_LIKE_SCAN_INTERVAL_HOURS, GRAPH_STORE_ACCOUNT_LIKE_MIN_VERSIONS_COUNT, and GRAPH_STORE_ACCOUNT_LIKE_MAX_UNIQUE_RATIO. (#6209)SubgraphIndexingStatus via the index-node API. (#6201)GRAPH_STORE_SETUP_TIMEOUT.GRAPH_STORE_DISABLE_CALL_CACHE, useful for indexers with locally hosted RPC nodes. (#6214)graph-node-debug image with frame pointers and linux-perf. (#6418)std::sync::RwLock with parking_lot::RwLock across multiple components for better performance under contention.ipfs.map() callbacks. (#6336) (#6416)nocase filter generation for non-String types (e.g., Bytes), which caused SQL errors like operator does not exist: bytea ~~* bytea. (#6351)can_copy_from failures not fully rolling back deployment creation, leaving orphaned records. (#6228)histogram_bounds statistics. (#6275)U256 in the runtime — now returns an error instead. (#6219)graphman copy copying pruned earliest_block_number from source, which could leave the destination deployment in an invalid state. (#6384)0x being incorrectly cached in the call cache, preventing stale empty results from persisting. (#6187)graphman dump and graphman restore (experimental). New commands for exporting and importing subgraph data in Parquet format, enabling backup and migration of deployment data across shards. Supports incremental dumps and progress reporting. See docs/dump.md for details. (#6397)graphman chain call-cache remove now supports --ttl-days to remove stale call cache entries that haven't been accessed within a specified number of days. (#6186)gnd as a drop-in replacement for graph-cli: init, add, codegen, build, publish, deploy, create, remove, auth, clean, and test commands. (#6282)Thanks to all contributors for this release: @DaMandal0rian, @dimitrovmaksim, @fubhy, @hudsonhrh, @incrypto32, @isum, @lutter, @shiyasmohd