Can any experts advise me. I’ve bought a 128GB SSD drive for my main computer (running OS 42.2 soon to be 42.3) which I’ll use for the root + swap + tmp files. I believe SSD drives both have a tendency to fail catastrophically and produce unrecoverable errors at a higher rate than “normal” HDs.
So, I’d like to mirror the SSD onto a part of the 2TB HD also installed.
so, some questions:
is partition level SW RAID 0 or RAID 1 is the best option, or is there a “whole disk” SW mirroring option. Please advise?
will this make setup / config / reconfig of the computer significantly more difficult / unreliable?
are there other considerations I should be aware of?
any specific recommendations (or references to setup guides) for this type of use?
I’ll be using LVM for the rest of the 2TB disk for the other partitions I need (for VMs).
SSD wearing as a danger for your data is from the beginning of SSD’s. My oldest one is ~8 years old, I thought it dies but a friend put it in another case+controller and it’s been doing fine since then. To get to the EOL for a modern SSD takes petabytes of writes.
Also the hardware you running on…what happens if that fails specifically the SATA controller, you might have one controller for both disks, if that fails, doesn’t help you RAID.
Does the system have two SATA controllers, if so, if you do go the RAID route you need to check each disk has is on it’s own controller. Else if this is a desktop, maybe a PCIe SATA controller will suffice.
Also I wonder how the I/O speed would be affected with an SSD and a HDD in a RAID setup, if anything SSD’s are used to cache the data feeding to a RAID setup for speed.
My suggestion, use the SSD for the OS and say /home (I use a /data and softlinks as it’s easier to backup), your 2TB device for LVM and get yourself a backup disk that will take your important data and store this somewhere offsite (I send mine to my wife’s office and sits in her bottom desk draw).
If you end up doing any kind of RAID, you can configure partition or whole disk.
Absolutely yes. RAID0 makes a system <very> vulnerable to any kind of file system failure by making your point of failure <anywhere> across 2 disks instead of only one. if disk I/O is your point of congestion, then RAID0 can address this, but is highly discouraged for anything other than writing log files.
RAID1 can improve certain fault tolerance at the cost of some complexity, it then becomes whether that complexity makes a diff to you. Probably little about setup, breaking an array, and re-building an array is difficult, but will you spend the time to familiarize? Most people don’t test and learn these procedures so it then becomes extra work without any benefit when a real problem happens. Remember that RAID1 only protects against some types of individual disk failures and nothing else.
Yes. As others have noted, your fear for “catastrophic” failure is likely baseless. On an SSD, if a hardware electrical path fails in a loss of memory chips, that would be catastropic and require hardware repair to maybe recover. But if you’re talking about cell/trap failures, those aren’t much different than when bad sectors go bad on HDD, and the disk firmware will then write that data to a different trap… similar to what would happen on a HDD. So, you’ll have to define “catastrophic failure” to determine whether your fear has any basis.
Thanks for all your replies. The consensus seems to be: mirroring not worth the effort and will decrease reliability & add complexity / opportunities for error on change. A good backup regime is a better use of my time & effort.
You should have that regardless of you using whatever.
While some benefits of both backups and RAIDs are common, IMHO they mainly address different things.
Backups to be able to restore older files. Varies from a user coming to you and admitting that he deleted something important by incident to a complete database that run bonkers and has to be restored to a certain well know point in time. No RAID will help here.
Disk arrays and other things behind it, like double power circuits, double data channels/buses, are to increase up-time.