Right now I have a single cache drive where I store my AppData Folder.  I'm wondering if that's the best option, since all the data written pretty much passes through the cache drive.  Is that dangerous since the cache drive, and therefor the System folder, doesn't have parity?  If the cache drive is lost, so is the System folder.  I do have the Community Application backup service running along with Crashplan, but I've always been funny about having a bunch of redundancy.

 

Perhaps I'm just being over cautious? 

 

Edit -> I just looked and mine is pretty old.  According to SMART data its been on for 5 years 9 months 

Edited August 10, 20205 yr by Spyderturbo007

2 minutes ago, Spyderturbo007 said:

Right now I have a single cache drive where I store my AppData Folder.  I'm wondering if that's the best option, since all the data written pretty much passes through the cache drive.  Is that dangerous since the cache drive, and therefor the System folder, doesn't have parity?  If the cache drive is lost, so is the System folder.  I do have the Community Application backup service running along with Crashplan, but I've always been funny about having a bunch of redundancy.

 

Perhaps I'm just being over cautious? 

It depends on what drive is the cache drive.

• HDD would be the riskiest as it is subjected to similar failure rate as other HDD in your array, which is relatively quite high.
• SATA SSD would be rather low risk. SSD rarely ever fails catastrophically (unlike HDD) so even if it's failing, you should have plenty of time to react.
  • The biggest risk to a single-drive cache pool with SATA SSD is disk dropping offline (e.g. SATA / power issue, controller issue etc.). If that happens during a write, you may end up with an unmountable pool due to corrupted file system. You can SOMEWHAT mitigate this risk by running btrfs with DUP profile for metadata (and system) chunks - depending on the SSD model and circumstances, it may give you a better chance of avoiding corrupted file system.
• NVMe SSD would have the lowest risk - I would personally consider it negligible. NVMe rarely (if ever) drops offline so you are only subjected to the typical low rate of failure for SSD.

 

 

17 minutes ago, testdasi said:

It depends on what drive is the cache drive.

• HDD would be the riskiest as it is subjected to similar failure rate as other HDD in your array, which is relatively quite high.
• SATA SSD would be rather low risk. SSD rarely ever fails catastrophically (unlike HDD) so even if it's failing, you should have plenty of time to react.
  • The biggest risk to a single-drive cache pool with SATA SSD is disk dropping offline (e.g. SATA / power issue, controller issue etc.). If that happens during a write, you may end up with an unmountable pool due to corrupted file system. You can SOMEWHAT mitigate this risk by running btrfs with DUP profile for metadata (and system) chunks - depending on the SSD model and circumstances, it may give you a better chance of avoiding corrupted file system.
• NVMe SSD would have the lowest risk - I would personally consider it negligible. NVMe rarely (if ever) drops offline so you are only subjected to the typical low rate of failure for SSD.

 

 

Thanks for the detailed response.  I currently have a 250GB SSD, which sometimes isn't enough so I've been considering upgrading it to 500GB or 1TB.  Would it be better to go with 2 x 500?  If I'm understanding the pool feature, it provides redundancy in the event of a cache drive failure, so I'm assuming either 2 x 250GB or 2x 500GB would be best depending on how much storage I think I'll need.

 

My server motherboard doesn't have an M.2 slot.

Edited August 10, 20205 yr by Spyderturbo007

1 minute ago, Spyderturbo007 said:

Thanks for the detailed response.  I currently have a 250GB SSD, which sometimes isn't enough so I've been considering upgrading it to 500GB or 1TB.  Would it be better to go with 2 x 500?  If I'm understanding the pool feature, it provides redundancy in the event of a cache drive failure, so I'm assuming either 2 x 250GB or 2x 500GB would be best depending on how much storage I think I'll need.

 

My server motherboard doesn't support an NVMe SSD.

The default profile for 2x drives in btrfs is RAID-1 which would provide redundancy so yes, 2x250 or 2x500 depends on your storage needs.

 

I do recommend you either use 6.9.0-beta25 or wait till 6.9.0-rc1 is out or wait till 6.9.0 stable is out before making the jump to multi-drive pool.

Given it's an SSD, you should have plenty of time to wait.

 

There's a bug with multi-drive btrfs pool that causes excessive write and/or poor performance and/or excessive CPU load with some SSD models.

This bug is fixed with 6.9.0 but it requires a clean format of the pool to align the partitions correctly (and consequently makes the newly-formatted pool incompatible with pre-6.9.0 versions).

 

 

 

Wow, thanks!  I'll wait for 6.9.0 to purchase the drives.

 

Thanks again for all your help testdasi!