Use the ebx.repository.directory property in the ebx.properties file to set the EBX® root directory location. This directory will contain the Lucene indexes and should be hosted on local, low latency hard disks. SSD disks are recommended. Using the local hard disk for indexes presents no risk of data loss, as EBX® is able to rebuild indexes from data stored in the relational database. However, the recovery operation can be time-consuming.
The EBX® root directory should never be shared by multiple EBX® instances.
You must avoid storing EBX® indexes over network attached volumes based on protocols, such as NFS, SMB, or SAMBA. The same restrictions apply to object-based storage (Amazon S3, Google Cloud Storage, Azure Blob) or file systems mounted over object storage (Amazon S3 file gateway). This may cause stability and performance issues; the more common symptoms are slow search speeds and running out of open files.
A high-performance Storage Area Network (SAN) can be an acceptable solution, especially if high performance technologies such as Fiber Channel (FC) are used.
On the cloud, EBX® has been tested with the following disks:
Amazon Web Services (AWS): Amazon Elastic Block Store (EBS) general purpose SSD volumes (gp3)
Microsoft Azure: Azure Standard SSD locally redundant storage (LRS)
Google cloud: Google Balanced persistent disks (pd-balanced)
Anti-virus software should not be used on the root directory as it can severely impact performance and can corrupt the index repository if a file is quarantined or removed. Exclusion by extension is not possible because the extensions used by Lucene can change at any time.
As a rule of thumb for the disk size, plan for 10 times the space occupied by the table G_BLK in the relational database.