Please let us configure swap allocation independent of pool layout
Description
One major change from FN 11.3 to TN Core 12.0 is the allocation of swap space, specifically the composition of gmirror devices upon boot.
In 11.3 FN created mirrored pairs regardless of the pool layout. This is the right thing to do, IMHO!
In 12.0 it tries to deduct the level of redundancy of the pool and configure the swap mirror in a way as to keep working with the maximum number of failed drives.
E.g. 4 drive RAIDZ2 pool --> 2 drives could fail and the pool would still be available --> a 3-way mirror is created for swap. Rendering the system with half of the possible swap space and one disk unused for swap.
A RAIDZ2 pool does not exist to keep it running with two failed disks. It exists to act as soon as a single disk fails and survive another failure during resilver. I can easily `swapoff` whatever devices I am working on. If two swap devices fail, who cares? It's just the swap, panic, reboot, data is still good ...
Please make this optional. I would highly appreciate it.
N.B. 2 GB of swap per disk is ridiculous. Why is this the default?
Kind regards, Patrick
SmartDraw Connector
Katalon Manual Tests (BETA)
Activity
Kris Moore
July 18, 2024 at 6:06 PM
Thank you for submitting this feature request! To better accommodate and gauge community interest for future versions of TrueNAS we have moved the submission process to our TrueNAS Community Forums. If this feature is still important and relevant for consideration, please refer to the links below on how to submit it for community voting and TrueNAS roadmap review.
Low
One major change from FN 11.3 to TN Core 12.0 is the allocation of swap space, specifically the composition of gmirror devices upon boot.
In 11.3 FN created mirrored pairs regardless of the pool layout. This is the right thing to do, IMHO!
In 12.0 it tries to deduct the level of redundancy of the pool and configure the swap mirror in a way as to keep working with the maximum number of failed drives.
E.g. 4 drive RAIDZ2 pool --> 2 drives could fail and the pool would still be available --> a 3-way mirror is created for swap. Rendering the system with half of the possible swap space and one disk unused for swap.
A RAIDZ2 pool does not exist to keep it running with two failed disks. It exists to act as soon as a single disk fails and survive another failure during resilver. I can easily `swapoff` whatever devices I am working on. If two swap devices fail, who cares? It's just the swap, panic, reboot, data is still good ...
Please make this optional. I would highly appreciate it.
N.B. 2 GB of swap per disk is ridiculous. Why is this the default?
Kind regards,
Patrick