Disk usage is unusually high or rapidly increasing on AEM Server
Learn how to identify the causes of high disk usage on AEM Server. Use a debug logger, capture thread dumps and do CPU profiling, and run the disk usage report.
Description description
Environment
Adobe Experience Manager
Issue/Symptoms
Disk usage is unusually high or rapidly increasing on an AEM server. With the disk space running out, AEM has stopped functioning.
Resolution resolution
A. If AEM is running and there is ample disk space
-
Configure oak write trace logging If AEM is still running then we can enable a debug logger to tell us which repository paths are being written to. To enable this logger, install the attached log configuration package below or follow these steps:
- Go to https://aemhost:port/system/console/slinglog
- Click on Add new logger.
- Configure a logger: Log File: logs/repgrowth.log, Log Level: trace, Loggers: org.apache.jackrabbit.oak.jcr.operations.writes
- Download file. This Package contains the required configuration for the logging write session for Oak. Install this package via the CRX Package Manager. After your monitor period, please make sure to uninstall the package.
Caution
- The log includes information regarding all writes and session details. If you use this logger, then make sure you have sufficient disk space.
- Uninstall the log configuration package or remove the log configuration after a short period of having this enabled to avoid further disk space consumption.
-
Run the disk usage report You can also leverage the Disk Usage report https://host:port/etc/reports/diskusage.html. This report displays the disk space used by repository path. The report is drillable, allowing you to view subtrees as well.
-
Capture thread dumps and perform profiling After using the repgrowth.log to get some idea of what data is being written, we can get information about what code is writing that data by capturing thread dumps and running CPU profiling. Visit these pages:
B. If AEM has stopped and/or disk space is almost full
If you had to stop AEM to avoid growing disk space, then use the commands below to do some initial analysis.
-
On Linux platform, leverage the du command to list all directories under crx-quickstart with the summarized size of those directories:
code language-none du -h --max-depth=2 crx-quickstart -
Use find and du commands to find recently modified files and get their sizes:
code language-none find crx-quickstart -type f -mtime 1 -exec du -hs {} \; -print -
To find large files in the datastore, you can combine find, du, and file commands to find files over 100MB in the datastore directory and auto-identify their file type:
code language-none find crx-quickstart/repository/datastore -type f -size +100M -exec sh -c "du -hs \"{}\"; file \"{}\"" \; -
If you find that growth is occurring in the segmentstore directory, then the command below might help indicate as to what data is being written:
code language-none strings data_xxxxxx.tar | egrep '.?/' | sed 's/.$//;s/.\//\//'
Cause
Some potential causes of unusual increases in disk utilization are:
- Proper maintenance hasn’t been run on the system. See this article for details on various system maintenance activities.
- AEM or the application is creating very large number of nodes or updates to node properties. This could be due to a misconfiguration or an application code bug. Since the tar storage in Oak operates in an append-only mode, repeated saving of nodes further contributes to excessive repository growth.
- Very large file(s) have been uploaded to AEM Assets or package manager.
- Debug or Trace logging was left enabled.