Skip to end of metadata
Go to start of metadata

Description

Geode JVMs use disk in several ways, including:

  • Logging and statistics
  • Region persistence and overflow
  • Gateway queue persistence and overflow
  • Client queue overflow

In an active system, these files (especially the region persistence files) can use a lot of disk space.

Determination

Disk Store Monitor

The Geode Disk Store Monitor runs in each JVM and monitors disk usage. For additional details, see the Apache Geode Documentation.

It logs a warning like below in the Geode log file when the disk usage exceeds the usage threshold (default=80%).

It throws errors and warnings like below when the disk usage exceeds the critical threshold (default=99%) and shuts down the JVM.

In conjunction with the Disk Store Monitor warnings and errors, an IOException with the message Not enough space left on device like below may also be thrown either by a Geode thread or an application thread.

df

One way to see disk space usage of a given disk is to use the operating system dfcommand like:


vsd

Another way to determine whether there is a disk issue is to use vsd to display the disk free space values contained in a given Geode statistics archive.

VMStats

The chart below shows DiskDirStats volumeFreeSpace values. The DiskDirStats volumeFreeSpace values show the available usable space in the directory. In this case, the disk store ran out of disk space. 

DiskDirStats

Action

There are several actions that can help alleviate disk issues issues, including:

  • Remove old Geode log and statistics files
  • Compact the Geode region and queue persistence files
  • Configure Geode log and statistics rolling so earlier logs and statistics files are deleted
  • Add additional disks for region and queue persistence and overflow

For additional details, see the Apache Geode Documentation.

  • No labels