Backing Up Memory Tier to Retain Storage Prior to Upgrade and Restart

Alluxio Worker storage can be mounted on memory, solid state, drives, or hard drives. Memory-mounted storage is fast, but volatile; by nature of it being file system mounted on memory, when the machine/VM is rebooted, the storage is lost.

The desire to keep data in memory cache may at times conflict with the need to perform maintenance or upgrades on the machine that require a restart. To ensure that memory storage "survives" a reboot, the following steps can be taken to ensure that memory-mounted cache can be stored to a durable location, and then recovered after reboot.

Assumptions:

1. Alluxio memory-mounted storage is mounted on /mnt/ramdisk
2. /some/tmp/location is durable storage - on either SSD or HDD

Steps to take:

• Stop the alluxio cluster:

alluxio-stop.sh all

• For each worker machine: Create a temporary location to store data from ramdisk:

mkdir /some/tmp/location

• Copy the contents of the memory-mounted storage to durable storage:

cp -R /mnt/ramdisk/* /some/tmp/location

Apply any patches and restart the machine:

alluxio-mount.sh SudoMount

Copy back the content of the ramdisk:

cp -R /some/tmp/location/* /mnt/ramdisk

Restart the alluxio cluster:

alluxio-start.sh all