You have to unzip the SPL, cd into the directory, run. Once the dependencies are installed, first you should compile the Solaris Porting Layer. You also need kernel-devel and a few other packages-this should work on a Red Hat-compatible distribution: yum groupinstall "Development Tools" yum install zlib-devel libuuid-devel libblkid-devel ↪libselinux-devel parted lsscsi You need two packages: the Solaris Porting Layer (this provides the Solaris kernel APIs) and the ZFS source. To get started with ZFS, first you need to download the source from. This means that no file can ever be “in the middle” of a write operation, so corruption leading to long fscks are greatly reduced. Copy-on-write (COW) means that when your data is read (for example, when you open a file) and then modified, a new copy is saved to disk, and the old copy is either re-allocated for future use or preserved as a “snapshot” of the current state of the filesystem. ZFS is notable because it's a mature implementation of a copy-on-write filesystem. (This, of course, taints your kernel's license.) This module is considered to be in release candidate state, but the core ZFS code itself is based on the fairly stable Solaris version. Recently, a group at Lawrence Livermore National Labs ported ZFS so that it's a separately compiled kernel module that can be installed on your system. Some people have ported ZFS to FUSE so that it can run in userspace and not taint the kernel, but this is slower than native kernel-space access. Unfortunately, Sun decided to release ZFS under the CDDL license, which ultimately has caused the kernel gurus never to add support to the main Linux kernel. ZFS has been something that many Linux users have desired for almost a decade because of its amazing features and simple-to-implement flexibility. The ZFS filesystem originally was created by Sun Microsystems for its Solaris operating system. If you then add to your system some off-site backups, you'll have a pretty good start to a robust backup solution. Assuming you create your storage pool on multiple disks in a mirror or RAID-Z configuration, you will have built-in protection against bitrot-the natural propagation of errors in your data from misreads or writes to disk. Then, I provide some code snippets you can use to create a rotating snapshot-like backup script using ZFS's native snapshot capability. I start with a quick overview of how to download and install the source code for this system on your Red Hat-compatible OS, and help you set up a small ZFS storage pool. You still can use hardware RAID if you want, but you're removing some of the built-in safeguards that help your filesystem prevent loss of data.įirst, I'm going to talk about ZFS. The usage of expensive hardware or slow software RAID systems is unnecessary and conflicts with how these filesystems expect to talk to storage disks. The biggest concept to grasp with ZFS and Btrfs is that ZFS and Btrfs expect disks to be disks. They also handle pools of storage in a fundamental way that makes an LVM layer less useful. The newer filesystems like ZFS and Btrfs (have or will soon) implement redundancy and internal checking for consistency without that extra layer of expensive RAID controllers or slow software. I also used LVM to make my storage more flexible and easier to manage. In the past, I relied on RAID hardware (and/or software) to handle drive redundancy and to increase performance, but that no longer is necessary. Disks are cheap, but not always reliable, so I've learned to adapt by using the latest generation of filesystems to overcome drive failures. I buy hard drives by the 20-pack, several times a year. I have to be-I run a small department and manage the backups of about ten different mission-critical servers.
0 Comments
Leave a Reply. |
AuthorWrite something about yourself. No need to be fancy, just an overview. ArchivesCategories |