With the release of Synex Server 13 R1, we introduced synex-zfs-installer, a native installer that enables ZFS infrastructure deployment without the limitations of the traditional Debian Installer. That first version laid the groundwork with single-disk installation support, but from the start we knew the next step was fundamental: data redundancy.
Why ZFS in enterprise environments
The choice of ZFS as the priority filesystem for Synex Server was not arbitrary. In environments where data integrity is critical, ZFS offers features that other filesystems simply cannot match natively.
Copy-on-write ensures that data is never partially overwritten: every write is atomic, eliminating the risk of corruption from power interruptions or system failures. Automatic checksums verify the integrity of every data block, silently detecting and correcting errors that in other systems would go unnoticed until it was too late. Transparent compression reduces storage consumption without administrator intervention. And instant snapshots enable recovery points with no performance cost.
But none of these features matter if the disk is lost. That is why redundancy is the natural complement to ZFS in production.
ZFS topologies: redundancy integrated into the filesystem
Unlike traditional RAID that operates at the block level beneath the filesystem, RAIDZ is integrated directly into ZFS. This architecture eliminates the problem known as “write hole” that affects RAID5/6, where an interruption during writing can leave the array in an inconsistent state.
We are implementing support for five topologies, from the simplest to highly redundant configurations:
- Mirror replicates data across two or more identical disks, equivalent to RAID1. Every write is duplicated to all drives in the mirror, providing full redundancy as long as at least one disk remains operational. It is the simplest option for redundancy and offers the best read performance, although with the highest cost in usable capacity.
- Stripe distributes data across multiple disks without redundancy, equivalent to RAID0. It maximizes capacity and performance by combining the space of all disks, but the loss of any drive means total data loss. Reserved for scenarios where performance is the priority and data can be rebuilt from other sources, such as caches or temporary storage.
- RAIDZ1 tolerates the loss of one disk, requiring a minimum of three drives. It is comparable to RAID5 in terms of redundancy, but without its inherent vulnerabilities. Suitable for environments where storage cost is a factor and the risk of simultaneous failure of two disks is acceptably low.
- RAIDZ2 tolerates the loss of two simultaneous disks, requiring a minimum of four drives. Recommended for high-capacity disks where rebuild time is prolonged, and where the probability of a second failure during reconstruction is not negligible.
- RAIDZ3 tolerates the loss of three disks, requiring a minimum of five drives. Designed for maximum criticality environments where even improbable failure scenarios must be considered.
Simplification from the installer
Configuring RAIDZ manually requires specific knowledge: correctly identifying devices, calculating pool geometry, executing commands with appropriate parameters, and ensuring the bootloader is correctly configured for a multi-disk system. It is not complex for an experienced administrator, but it is fertile ground for errors under pressure.
The work we are doing integrates this entire process into the installation flow. The administrator selects the desired topology, the installer validates that the number of disks is sufficient, presents available devices for interactive selection, and automatically manages partitioning, pool creation, and boot configuration.
The goal is for deploying a server with RAIDZ3 across five disks to be as straightforward as installing on a single disk: select, confirm, wait. Without sacrificing control for those who need to adjust specific parameters, but without requiring that knowledge for those who simply need functional redundancy.
Next steps
This functionality will be part of Synex Server 13 R2. In upcoming posts, we will share progress on other improvements in development, including encryption options and support for additional filesystems.
Synex Server 13 R1 is available for download from here.