Docker下使用neo4j-admin导入数据后容器无法启动的问题求助

问题场景 #

执行以下Docker命令运行Neo4j 5.11.0社区版并导入CSV数据：

docker run --publish=7474:7474 --publish=7687:7687 \
--volume=C:/mydir/import:/import \
--env NEO4J_AUTH=neo4j/password \
neo4j:5.11.0-community \
bin/neo4j-admin database import full --id-type=INTEGER --nodes=NodeA=/import/nodesa.csv --nodes=NodeB=/import/nodesb.csv --relationships=RELA=/import/rels.csv --multiline-fields=true

导入日志显示执行成功：

IMPORT DONE in 13s 945ms.
Imported:
 75650 nodes
 95438 relationships
 255509 properties
Peak memory usage: 547.9MiB

但Docker Desktop中容器并未处于运行状态，尝试启动容器时触发报错：

java.lang.RuntimeException: java.nio.file.DirectoryNotEmptyException: /data/databases/neo4j already contains data, cannot do import here

重启容器后的完整日志：

2023-10-05 19:47:47 Neo4j version: 5.11.0
2023-10-05 19:47:47 Importing the contents of these files into /data/databases/neo4j:
2023-10-05 19:47:47 Nodes:
2023-10-05 19:47:47   [NodeA]:
2023-10-05 19:47:47   /import/nodea.csv
2023-10-05 19:47:47 
2023-10-05 19:47:47   [NodeB]:
2023-10-05 19:47:47   /import/nodeb.csv
2023-10-05 19:47:47 
2023-10-05 19:47:47 Relationships:
2023-10-05 19:47:47   RELA:
2023-10-05 19:47:47   /import/rels.csv
2023-10-05 19:47:47 
2023-10-05 19:47:47 
2023-10-05 19:47:47 Available resources:
2023-10-05 19:47:47   Total machine memory: 11.70GiB
2023-10-05 19:47:48   Free machine memory: 10.06GiB
2023-10-05 19:47:48   Max heap memory : 2.601GiB
2023-10-05 19:47:48   Max worker threads: 8
2023-10-05 19:47:48   Configured max memory: 6.760GiB
2023-10-05 19:47:48   High parallel IO: true
2023-10-05 19:47:48 
2023-10-05 19:47:52 
2023-10-05 19:47:52 IMPORT FAILED in 0ms. 
2023-10-05 19:47:52 Data statistics is not available.
2023-10-05 19:47:52 Peak memory usage: 0B
2023-10-05 19:47:52 Database already exist. Re-run with `--overwrite-destination` to remove the database prior to import
2023-10-05 19:47:52 Caused by:/data/databases/neo4j already contains data, cannot do import here
2023-10-05 19:47:52 
2023-10-05 19:47:52 WARNING Import failed. The store files in /data/databases/neo4j are left as they are, although they are likely in an unusable state. Starting a database on these store files will likely fail or observe inconsistent records so start at your own risk or delete the store manually
2023-10-05 19:47:52 java.lang.RuntimeException: java.nio.file.DirectoryNotEmptyException: /data/databases/neo4j already contains data, cannot do import here
2023-10-05 19:47:52 at org.neo4j.importer.CsvImporter.andPrintError(CsvImporter.java:358)
2023-10-05 19:47:52 at org.neo4j.importer.CsvImporter.doImport(CsvImporter.java:268)
2023-10-05 19:47:52 at org.neo4j.importer.CsvImporter.doImport(CsvImporter.java:181)
2023-10-05 19:47:52 at org.neo4j.importer.ImportCommand$Base.doExecute(ImportCommand.java:405)
2023-10-05 19:47:52 at org.neo4j.importer.ImportCommand$Full.execute(ImportCommand.java:531)
2023-10-05 19:47:52 at org.neo4j.cli.AbstractCommand.call(AbstractCommand.java:92)
2023-10-05 19:47:52 at org.neo4j.cli.AbstractCommand.call(AbstractCommand.java:37)
2023-10-05 19:47:52 at picocli.CommandLine.executeUserObject(CommandLine.java:2041)
2023-10-05 19:47:52 at picocli.CommandLine.access$1500(CommandLine.java:148)
2023-10-05 19:47:52 at picocli.CommandLine$RunLast.executeUserObjectOfLastSubcommandWithSameParent(CommandLine.java:2461)
2023-10-05 19:47:52 at picocli.CommandLine$RunLast.handle(CommandLine.java:2453)
2023-10-05 19:47:52 at picocli.CommandLine$RunLast.handle(CommandLine.java:2415)
2023-10-05 19:47:52 at picocli.CommandLine$AbstractParseResultHandler.execute(CommandLine.java:2273)
2023-10-05 19:47:52 at picocli.CommandLine$RunLast.execute(CommandLine.java:2417)
2023-10-05 19:47:52 at picocli.CommandLine.execute(CommandLine.java:2170)
2023-10-05 19:47:52 at org.neo4j.cli.AdminTool.execute(AdminTool.java:94)
2023-10-05 19:47:52 at org.neo4j.cli.AdminTool.main(AdminTool.java:82)
2023-10-05 19:47:52 Caused by: java.nio.file.DirectoryNotEmptyException: /data/databases/neo4j already contains data, cannot do import here
2023-10-05 19:47:52 at org.neo4j.internal.batchimport.store.BatchingNeoStores.assertDatabaseIsNonExistent(BatchingNeoStores.java:230)
2023-10-05 19:47:52 at org.neo4j.internal.batchimport.store.BatchingNeoStores.createNew(BatchingNeoStores.java:203)
2023-10-05 19:47:52 at org.neo4j.internal.batchimport.ParallelBatchImporter.doImport(ParallelBatchImporter.java:131)
2023-10-05 19:47:52 at org.neo4j.importer.CsvImporter.doImport(CsvImporter.java:264)
2023-10-05 19:47:52 ... 15 more

问题原因 #

• 容器执行逻辑问题：docker run命令后指定neo4j-admin import时，容器会仅执行该导入任务，任务完成后直接退出，不会自动启动Neo4j服务，因此容器处于停止状态。
• 重复导入冲突：重启容器时，Docker会重复执行启动命令中的neo4j-admin import，但第一次导入的数据已保存在/data/databases/neo4j目录，而neo4j-admin import full默认不允许覆盖已有数据库，因此触发报错。
• 与社区版无关：该限制是neo4j-admin import工具的通用规则，社区版和企业版均遵循此逻辑。

解决方法 #

方法1：分步骤执行导入与启动（推荐） #

将导入和启动服务拆分为两个独立命令，避免重复执行导入：

1. 执行导入任务，挂载数据卷保存结果，任务完成后自动删除临时容器：

docker run --rm \
--volume=C:/mydir/import:/import \
--volume=C:/mydir/neo4j-data:/data \
neo4j:5.11.0-community \
bin/neo4j-admin database import full --id-type=INTEGER --nodes=NodeA=/import/nodesa.csv --nodes=NodeB=/import/nodesb.csv --relationships=RELA=/import/rels.csv --multiline-fields=true

2. 启动正常的Neo4j容器，挂载保存了导入数据的卷：

docker run -d --publish=7474:7474 --publish=7687:7687 \
--volume=C:/mydir/neo4j-data:/data \
--env NEO4J_AUTH=neo4j/password \
neo4j:5.11.0-community

方法2：单命令组合导入与启动 #

通过bash -c将导入命令和启动命令组合成一个容器启动命令：

docker run -d --publish=7474:7474 --publish=7687:7687 \
--volume=C:/mydir/import:/import \
--env NEO4J_AUTH=neo4j/password \
neo4j:5.11.0-community \
bash -c "bin/neo4j-admin database import full --id-type=INTEGER --nodes=NodeA=/import/nodesa.csv --nodes=NodeB=/import/nodesb.csv --relationships=RELA=/import/rels.csv --multiline-fields=true && bin/neo4j start"

注意：此方法仅适合首次执行，若容器重启会再次执行导入命令，仍会触发冲突，建议优先使用方法1。

方法3：覆盖已有数据库重新导入 #

如果需要重新导入数据，可在neo4j-admin import命令中添加--overwrite-destination参数，该参数会删除已有数据库后重新导入：

docker run --rm \
--volume=C:/mydir/import:/import \
--volume=C:/mydir/neo4j-data:/data \
neo4j:5.11.0-community \
bin/neo4j-admin database import full --overwrite-destination --id-type=INTEGER --nodes=NodeA=/import/nodesa.csv --nodes=NodeB=/import/nodesb.csv --relationships=RELA=/import/rels.csv --multiline-fields=true

执行前请确认已有数据可被覆盖。

内容的提问来源于stack exchange，提问作者iuhztr tzhjzuk