For VM performance and backup optimization I use separate disks for temporary data like:
- swap
- logs
- system temp-folders
- temporary DBs (like tmpdb in MSSQL)
- local backups, software packages and many other not-critical stuff
Separate disks allow 4 important things:
1) Efficient disk-cache settings - "unsafe" is perfect for such data
2) Unsafely-optimized host storage type or place (for example, ext4 with disabled journaling; partitions that put out from RAID)
* Also, I was going to use RAW instead of QCOW2, but for now it's not possible if VM-snapshots are necessary (even if I don't need snapshots for these drives)
3) Unsafely-optimized guest FS-settings or FS-type (like no-journaled exFAT instead of NTFS for some cases)
4) Much faster backups and less space usage with excluding such disks from backups
It makes VMs faster, while saving host and PBS resources (surprise to not see something like this in stock performance recommendations)
However, there are 2 problems for restore such VMs with no-backup drives:
1. Drives are deleted from VM-config. It makes restore-process longer and demands manual intervention.
2. The very genius feature "Live restore" can't be used at all when such disk may be empty, but necessary to exists.
Because if I manually put drives in config during restore, it just fails.
Are there any solutions or maybe already proposed features to track?
I think it might be something like:
1) "Unreferenced disk" as optional template for replacing no-exist disk
2) Config-override or at least "save original" flag before restore-process
3) Create and prepare disk -> single backup -> then exclude from backup + restore-option to use "disk image from older backup if it's not exist in current"
4) "Out-of-the-box" separate device-kind (like for EFI-disks) with 1 partition, list of predefined FS and "never-keep"\"local-keep" options
- swap
- logs
- system temp-folders
- temporary DBs (like tmpdb in MSSQL)
- local backups, software packages and many other not-critical stuff
Separate disks allow 4 important things:
1) Efficient disk-cache settings - "unsafe" is perfect for such data
2) Unsafely-optimized host storage type or place (for example, ext4 with disabled journaling; partitions that put out from RAID)
* Also, I was going to use RAW instead of QCOW2, but for now it's not possible if VM-snapshots are necessary (even if I don't need snapshots for these drives)
3) Unsafely-optimized guest FS-settings or FS-type (like no-journaled exFAT instead of NTFS for some cases)
4) Much faster backups and less space usage with excluding such disks from backups
It makes VMs faster, while saving host and PBS resources (surprise to not see something like this in stock performance recommendations)
However, there are 2 problems for restore such VMs with no-backup drives:
1. Drives are deleted from VM-config. It makes restore-process longer and demands manual intervention.
2. The very genius feature "Live restore" can't be used at all when such disk may be empty, but necessary to exists.
Because if I manually put drives in config during restore, it just fails.
Are there any solutions or maybe already proposed features to track?
I think it might be something like:
1) "Unreferenced disk" as optional template for replacing no-exist disk
2) Config-override or at least "save original" flag before restore-process
3) Create and prepare disk -> single backup -> then exclude from backup + restore-option to use "disk image from older backup if it's not exist in current"
4) "Out-of-the-box" separate device-kind (like for EFI-disks) with 1 partition, list of predefined FS and "never-keep"\"local-keep" options