What i have been doing is anything over xx amount ill disable cache on that share and dump the data straight to the array and switch the cache back on after. Most the time i only write a few GB to the cache a day so not a major deal but if i can figure out why its doing what its doing thats even better.

 

Previously I had both cache drives as the default raid 1 and i thought maybe raid 0 would be different but it seems not. When everything stalls out unraid all my dockers and the transfer become unresponsive until either i pause the transfer or let it complete

 

all shares are set to use cache and both ssds are the same,

 

 

Edited June 4, 20179 yr by sourCream

Do you regularly trim the pool?

That i do, it runs daily at 3am right after the mover at 1am

Post your diags, make sure they include at least one trim operation.

Just ran a trim.

atlantis-diagnostics-20170604-1547.zip

 

Jun 4 15:45:01 Atlantis root: /etc/libvirt: 920.7 MiB (965451776 bytes) trimmed
Jun 4 15:45:01 Atlantis root: /var/lib/docker: 15.3 GiB (16398581760 bytes) trimmed
Jun 4 15:45:01 Atlantis root: /mnt/cache: 12 GiB (12885016576 bytes) trimmed

 

shouldn't the 12GB be more like 512GB if i have 2 256GB in raid 0?

 

Ive got two of these http://www.ebay.com/itm/Micron-RealSSD-C400-2-5-256GB-SATA-6Gb-s-MTFDDAK256MAM-1K1-SSD/152496932175

Edited June 4, 20179 yr by sourCream

That's normal with btrfs, it's the actual trimmed bytes.

 

Don't see nothing out of the ordinary, trim is working and despite those not being the fastest SSDs they should be a lot faster than that.

 

Once my cache became inexplicably slow, re-formatting and starting over brought the speed back to what it should be, maybe worth a try.

Ya i thought about that. Worth a shot. I have a backup of the cache drive and my VM so i can reformat both drives. How should i go about it tho do i need to stop the array make a new config and tell it not to touch any of the array disks and just remove the cache drives?

If you want you can use the mover to backup and restore the cache contents, see here:

 

Then instead of replacing the cache just stop the array and wipe both SSDs:

 

blkdiscard /dev/sdX

 

After wiping both start the array and you'll have the option to format the pool, finally restore the data.

So essentially just move all the data to the array and move it back when done that way nothing changes and i dont have to pull down my dockers again? 

 

Where do i enter that command at and im assuming sdX is the actual drive letter of the drive i wish to format

14 minutes ago, sourCream said:

So essentially just move all the data to the array and move it back when done that way nothing changes and i dont have to pull down my dockers again? 

 

Correct

 

14 minutes ago, sourCream said:

Where do i enter that command at and im assuming sdX is the actual drive letter of the drive i wish to format

 

It won't format, just wipe, format is done after array start, enter the commands using the console/SSH.

Awesome thank you once again for the help. Will definitely give it a shot.