Skip to main content

BI Server Disk Partitioning Recommendations

Background

Three data disks are currently recommended: one application disk, one data disk, and one backup disk. Multiple disks are used so disk I/O does not interfere with each other during runtime, improving performance. This also avoids situations where, for example, poorly written SQL extracts too much dataset data and causes the entire business system to fail. After disks are separated, even if the data disk becomes full, users can still log in to the business system and continue accessing some pages.

Disk Partition Contents

DiskPathPurpose
Application diskSSD cloud disk, 300 GB or above. Mount path: /home/guandata/dockerBI runtime environment, images, and Spark shuffle. This disk stores Docker data, including Docker images, Docker logs, and temporary data in each running Docker container. When tasks run, Spark shuffle files are stored here, namely dataset and workflow runtime data. After a task finishes, the data is written to directories under /home/guandata/, and intermediate data in Docker is released. It does not occupy a large amount of space long term, but large workflows can consume significant space during runtime, so the disk size should not be reduced to avoid filling the disk during large workflows.
Data diskSSD cloud disk, 400 GB or above. Mount point path: /home/guandataBI data and cache. Stores actual data.
Backup diskStandard cloud disk, 300 GB or above. Mount path: /home/guandata/backup_dataStores BI backup data. The default backup time is 02:30.
Note

In some cases, specific paths may have been required during BI installation. Confirm the actual paths based on the environment. The paths above are the default paths.