[CASSANDRA-19401] Nodetool import expects directory structure

• Type: Bug
• Status: Resolved
• Priority: Normal
• Resolution: Fixed
• Fix Version/s: 4.0.13, 4.1.5, 5.0-rc1, 5.1-alpha1
• Component/s: Local/SSTable
• Bug Category: Correctness - Recoverable Corruption / Loss
• Severity: Normal
• Complexity: Normal
• Discovered By: User Report
• Platform: All
• Since Version: 4.0.0

Source Control Link

https://github.com/apache/cassandra/commit/334ca05730d8088f27323166e3ee1505db5cb202

Test and Documentation Plan

CI

• Assignee: Stefan Miklosovic
• Reporter: Norbert Schultz

Authors

Stefan Miklosovic

Reviewers

Brandon Williams

• Created: 15/Feb/24 09:48
• Updated: 02/May/24 09:24
• Resolved: 02/May/24 08:56

Description

According to the documentation the nodetool import should not rely on the folder structure of the imported sst files:

Because the keyspace and table are specified on the command line for nodetool import, there is not the same requirement as with sstableloader, to have the SSTables in a specific directory path. When importing snapshots or incremental backups with nodetool import, the SSTables don’t need to be copied to another directory.

However when importing old cassandra snapshots, we figured out, that sstables still need to be in a directory called like $KEYSPACE/$TABLENAME files, even when keyspace and table name are already present as parameters for the nodetool import call.

Call we used:

nodetool import --copy-data mykeyspace mytable /full_path_to/test1

Log:

INFO  [RMI TCP Connection(21)-127.0.0.1] 2024-02-15 10:41:06,565 SSTableImporter.java:72 - Loading new SSTables for mykeyspace/mytable: Options{srcPaths='[/full_path_to/test1]', resetLevel=true, clearRepaired=true, verifySSTables=true, verifyTokens=true, invalidateCaches=true, extendedVerify=false, copyData= true}
INFO  [RMI TCP Connection(21)-127.0.0.1] 2024-02-15 10:41:06,566 SSTableImporter.java:173 - No new SSTables were found for mykeyspace/mytable

However, when we move the sstables (.db-Files) to alternative/mykeyspace/mytable

and import with

nodetool import --copy-data mykeyspace mytable /fullpath/alternative/mykeyspace/mytable

the import works

INFO  [RMI TCP Connection(23)-127.0.0.1] 2024-02-15 10:43:36,093 SSTableImporter.java:177 - Loading new SSTables and building secondary indexes for mykeyspace/mytable: [BigTableReader(path='/mnt/ramdisk/cassandra4/data/mykeyspace/mytable-561a12d0cbe611eead78fbfd293cee40/me-2-big-Data.db'), BigTableReader(path='/mnt/ramdisk/cassandra4/data/mykeyspace/mytable-561a12d0cbe611eead78fbfd293cee40/me-1-big-Data.db')]
INFO  [RMI TCP Connection(23)-127.0.0.1] 2024-02-15 10:43:36,093 SSTableImporter.java:190 - Done loading load new SSTables for mykeyspace/mytable

We experienced this in Cassandra 4.1.3 on Java 11 (Linux)

Issue Links

• contains: CASSANDRA-19572 Test failure: org.apache.cassandra.db.ImportTest flakiness
• relates to: CASSANDRA-19572 Test failure: org.apache.cassandra.db.ImportTest flakiness
• links to: GitHub Pull Request #3259
• links to: GitHub Pull Request #3260
• links to: GitHub Pull Request #3261
• links to: GitHub Pull Request #3264