logo
Contact Enterprise Sales

Release-notes

NebulaGraph v3.0.0 Release Note

NebulaGraph v3.0.0 Release Note

NebulaGraph v3.0.0 is here! The new version introduced a series of new features including enhanced backup and restore, better support for the openCypher query language, and more fine-grained user management.

Features

  • Support backup and restore. #3469 #1 #22

  • Support openCypher queries with multiple MATCH statements. #3519 #3318

  • Support Standalone NebulaGraph. #3310

  • Support key value separation for the storage engine. #3281

  • Support topN push down for LOOKUP. #3499

  • Support vertex without tag. #3316 #3335 #3328 #3286

  • Support parameterized queries. #3379

  • Support queries without specifying VIDs but a LIMIT clause must be used to restrict the number of results. #3320 #3329 #3262

  • Support duration. #3338

  • Support most UTF-8 encoded characters of 1 to 4 bytes in Schema. #3380 #3440

  • Support DESCRIBE USER. #3300

Optimizations

  • Refactor cluster management. #3343

  • Add log monitor to check free bytes for log disks, change log level when your graph space is almost full. #3576

  • Support any string for tag names in apostrophe. #3424

  • Support that the storage service sends partition disk paths to the meta. #3369 #3416

  • Add constraints on invalid password attempts. #3573 #3629

  • Support DELETE in TOSS. #3374

  • Support to use logrotate. #3541

  • Support more metrics. #3446 #3605 #3590

  • Enhancement date and time parser. #3179

  • Remove read lock in the meta service to reduce the side effect of read-write locks. #3256

  • Refactor storage indexes to solve the coupling problem between services. #3196

  • Support specifying the floating point accuracy of the round() function. #3178ß

  • Support HTTPS for the ES client. #3150

  • Move version info outside of heartbeat. #3378

  • Support empty list, set, and map. #3302

  • Support specifying s2 region coverage parameters when creating a geo index. #3396

  • Add version info for SHOW HOSTS. #3702

Bugfix

  • Fix the bug that memory is not released when a default value is used and no value is specified in nGQL. #3666

  • Fix the bug that the function coalesce() cannot be used. #3653

  • Fix the bug that using multiple INSERT on an indexed tag will lead to incorrect LOOKUP results. #3627

  • Fix the crash when the expression exceeds the depth. #3606

  • Disable the aggregate function in the YIELD and WHERE clauses of nGQL. #3597

  • Fix the crash when using the aggregate function in UNWILD and WHERE clauses. #3397 #3355

  • Fix the bug that the tag index is rebuilt with an old schema version value. #3332

  • Fix the bug that the query results will still contain the expired edges if we use GO...REVERSELY. #3536

  • Fix the memory estimation error info in CentOS 6.0. #3534

  • Fix the crash when the LOOKUP statement contains a filter that consists of a logical And expression and an IN expression with only one element. #3525

  • Fix the bug that metad progress is suspended under heavy load. #3482

  • Fix the unwinding subgraph crash. #3506

  • Fix the DROP SPACE crash when rebuilding an index. #3406

  • Fix the bug of reading memory stats under cgroup v2. #3419

  • Fix the bug that DROP TAG INDEX deletes the edge index with the same name unexpectedly, and vice versa for the deletion of the tag index. #3413

  • Fix the bug that edges are not shown after a graph space is cloned. #3351

  • Fix the index existence check problem. #3315

  • Fix a bug that running the ALTER statement to query the type property may lead to a null pointer obtained by the graph service. #3325

  • Optimized the Raft to make the system more stable. #3172 #3435 #3358 #3322 #3031

  • Cancel memory check when the ratio is greater than 1.0. #3289

  • Fix the error with ninja-build. #3195

  • Fix the bug that creating tag and edge with the same name at the same time may both succeed. #3735

  • Fix the full-text index creation failure when there are the same tags or edges in different spaces. #3747

  • Fix variable inconsistency in YIELD clauses and GO statements. #3430

  • Fix the crash when schema version is greater than 256. #3893

Incompatibility

  • NebulaGraph 3.0.0 does not support most ecosystem tools of v2.x, please upgrade the ecosystem tools.

  • Storage services defined in the configuration files cannot be read or written directly. The configuration files only register the storage services into the meta services. You must run the ADD HOSTS command to read and write data on storage servers. #3343

  • Disable ZONE and GROUP. #3776 #3825 #3330

  • Disable BALANCE DATA. #3756

  • Modify the default session timeout from 0 to 28800 seconds, limit the value to between 1 and 604800 seconds. #3357 #3807

  • Add SHOW LOCAL SESSIONS and SHOW LOCAL QUERIES commands, and deprecate SHOW ALL QUERIES. #3488

  • A tag is not required for a vertex. DELETE VERTEX only deletes the vertices and does not delete the related outgoing and incoming edges of the vertices. At this time, there will be hanging edges by default. #3316 #3335 #3328 #3286

  • Disable the YIELD clause to return custom variables. #3271

  • The YIELD clause is required in the FETCH, GO, LOOKUP, FIND PATH and GET SUBGRAPH statements. #2957 #3056 #3139

  • Add non-reserved keywords: s2_max_level, s2_max_cells. #3396

  • It is required to specify a tag to query properties of a vertex in a MATCH statement. For example, from return v.name to return v.player.name. #3255

If you encounter any problems in the process of using NebulaGraph, please refer to NebulaGraph Database Manual to troubleshoot the problem. It records in detail the knowledge points and specific usage of the graph database and the graph database NebulaGraph.

Join our Slack channel if you want to discuss with the rest of the NebulaGraph community!