Strange behaviour with tables (hashtag2775)

[vikramsubramanian]

The following queries cause some weird behaviour with kuzu

create node table Person(id serial, primary key(id));
create rel table Knows(FROM Person TO Person);
create (:Person {id: 0});

From the last query, I get a strange error message: Error: Cannot write to file. path: tmpdatabase/data.kz fileDescriptor: 7 offsetToWrite: 17592186040320 numBytesToWrite: 4096 numBytesWritten: 18446744073709551615. Error: File too large When dropping the table:

drop table Knows;

Kuzu hangs. I've tried this both on my local computer and on the servers. )

[mayil-ai[bot]]

Summary: Strange behavior with table creation and dropping causing error messages and hanging in Kuzu.

Possible Solution

Based on the error message provided, the issue seems to be related to a file system limitation where the file size exceeds the maximum allowed by the system. The error message indicates an attempt to write to a file at a very large offset (offsetToWrite: 17592186040320), which suggests that the file has grown too large.

To resolve the issue:

• Check the file system limits for maximum file size. If you are on a 32-bit system or using a file system with a 2GB or 4GB file size limit, consider moving to a 64-bit system or a file system that supports larger files, such as NTFS or ext4.
• If the file system and operating system support large files, ensure that the application is compiled with large file support. In C++, this can typically be done by defining _FILE_OFFSET_BITS=64 or using the -D_FILE_OFFSET_BITS=64 compiler flag.
• Investigate the database's file handling code to ensure it properly supports large files. This may involve using 64-bit data types for file offsets and sizes.
• Consider implementing a mechanism to split the database across multiple files if a single file cannot be extended further due to file system limitations.
• Review the database's data storage and compaction strategies to prevent unnecessary growth of the database file. This could involve more efficient data encoding, periodic compaction, or archiving of old data.
• If the database supports it, consider configuring it to use multiple smaller files rather than a single large file. This could involve sharding or partitioning the data.
• Ensure that the database and the file system are properly configured to handle the expected data volume without hitting file size limits.

Code snippets to check

• src/catalog/catalog_content.cpp

  The issue may be related to how the catalog handles the creation and dropping of tables, which could be causing the strange error and hang.

• src/binder/bind/bind_query.cpp

  The issue might be related to how queries are bound and processed, especially since the error occurs on a query execution.

• src/binder/bind/bind_copy.cpp

  The COPY command is used in the test code, and if the issue is related to copying data into tables, this file might contain relevant code.

• src/binder/query/query_graph_label_analyzer.cpp

  The error message could be related to how the query graph labels are analyzed, which is part of the query processing.

• examples/c/main.c

  This file contains code that creates a node table and executes queries, which is similar to the issue described.

• examples/cpp/main.cpp

  This file contains code that creates a node table and executes queries, which is similar to the issue described.

• test/copy/e2e_copy_transaction_test.cpp

  This test file includes commands for creating and copying tables, which might be related to the issue if the problem is with the COPY command.

• tools/java_api/test.java

  This file includes Java API usage for creating and querying tables, which might be relevant if the issue is with the Java API.

• tools/java_api/src/test/java/com/kuzudb/test/QueryResultTest.java

  This file contains tests for query results, which might be relevant if the issue is with error handling in query results.