Java Heap when export data of postgres

General Discussion
1

I’m trying to export data from a Postgres database to use as seeds for other databases.

We configured this database to be our “model database.” Whenever a developer needs to upload a dynamic database to test their development, this data will be imported.

However, we’re getting Java heap errors. We’ve tested many machine configurations and Java memory parameters, and we always end up with heap errors.

The following command was executed:

export JAVA_OPTS="-Xmx24g -Xms24g -Xss2m -XX:+UseG1GC -XX:MaxGCPauseMillis=200 -XX:+UnlockExperimentalVMOptions -XX:+UseStringDeduplication"
./liquibase \
  --url="jdbc:postgresql://IP:PORT/postgres" \
  --username=postgres \
  --password=Postgres16 \
  --changeLogFile=equalizer/changelog/versions/222.xml \
  --diffTypes="data" \
  --excludeObjects="table1,table2,table3,table4" \
  --logLevel=fine \
  --logFile=liquibase.log \
  generateChangeLog

The database is considerably small, being 4408MB, 4.126 tables and 96 views.

I think this shouldn’t be a problem for Liquibase considering we have plenty of machine resources.

Liquibase version 4.33

JDK 17

2

Add: When trying to run export as SQL:

[2025-09-01 11:56:17] INFO [liquibase.command] Command execution complete
[2025-09-01 11:56:17] FINE [liquibase.servicelocator] Loaded liquibase.analytics.AnalyticsListener instance liquibase.analytics.LiquibaseAnalyticsListener
[2025-09-01 11:56:17] FINE [liquibase.servicelocator] Loaded liquibase.analytics.AnalyticsListener instance liquibase.analytics.NoOpAnalyticsListener
[2025-09-01 11:56:17] FINE [liquibase.configuration] No configuration value for liquibase.analytics.devOverride found
[2025-09-01 11:56:17] FINE [liquibase.configuration] Configuration liquibase.analytics.devOverride is using the default value of false
[2025-09-01 11:56:17] FINE [liquibase.configuration] No configuration value for liquibase.analytics.configEndpointUrl found
[2025-09-01 11:56:17] FINE [liquibase.configuration] Configuration liquibase.analytics.configEndpointUrl is using the default value of https://config.liquibase.com/analytics.yaml
[2025-09-01 11:56:17] FINE [liquibase.configuration] No configuration value for liquibase.analytics.enabled found
[2025-09-01 11:56:17] FINE [liquibase.configuration] No configuration value for liquibase.analytics.configCacheTimeoutMillis found
[2025-09-01 11:56:17] FINE [liquibase.configuration] Configuration liquibase.analytics.configCacheTimeoutMillis is using the default value of 3600000
[2025-09-01 11:56:17] FINE [liquibase.servicelocator] Loaded liquibase.analytics.configuration.AnalyticsConfiguration instance liquibase.analytics.configuration.LiquibaseRemoteAnalyticsConfiguration
[2025-09-01 11:56:17] FINE [liquibase.configuration] No configuration value for liquibase.analytics.logLevel found
[2025-09-01 11:56:17] FINE [liquibase.configuration] Configuration liquibase.analytics.logLevel is using the default value of OFF
[2025-09-01 11:56:17] FINE [liquibase.configuration] No configuration value for liquibase.analytics.configEndpointTimeoutMillis found
[2025-09-01 11:56:17] FINE [liquibase.configuration] Configuration liquibase.analytics.configEndpointTimeoutMillis is using the default value of 1500
[2025-09-01 11:56:19] FINE [liquibase.configuration] No configuration value for liquibase.license.utility.enabled found
[2025-09-01 11:56:19] FINE [liquibase.configuration] Configuration liquibase.license.utility.enabled is using the default value of false
[2025-09-01 11:56:19] SEVERE [liquibase.integration] Required array length 2103428988 + 56955090 is too large
java.lang.OutOfMemoryError: Required array length 2103428988 + 56955090 is too large
3

The problem was solved by splitting the export into three files.

However, I really think Liquibase should be able to handle situations like this.

4

Out of curiosity, what was the size of the 3 files and or how many rows of data were exported? The load and export data capabilities in Liquibase were designed to handle small volumes necessary for unit testing, integration testing and the management of concise sets of static lookup data. I’ve fielded several questions recently about larger data loads and I am trying to better understand the use cases and volumes that users need to work with.