Hi,

I am questioning my actual SSD pools configuration and I am wondering if a should plan for a change.

 

My setup is with 2 pools with a RAID1 of 2 500GB RED SSD each:

• cache (oscillating between 30 to 300GB with): for general caching of most of my shares. I also have some cache prefer shares with data that I want to edit before moving it to manually to a cached share.
• local_data (around 230GB): for my docker appdata (90%), domains, isos, system (9%) and nextcould

 

I do not do any downloading on the server, I manage all the file transfers manually so I do not risk having a pool filling up if I am not careful.

 

This works well but there are three points that make me think about only using a single pool with the 4 drives and keep the same level of protection:

1. One pools means I would not have to be AS careful as I am actually.
2. I might get a bit more speed from the pool this way.
3. I wonder if splitting the appdata over 4 drives would spread the writes better and improve the longevity overall.

 

Maybe I am overthinking it but an outside opinion is welcome.

Here is some more infos:

 

 

 

 

Single raid10 pool should give you better performance and also better distribute the writes among all devices, on the flip side any pool issue will affect all that data, but as long as you have backups I would probably go with a single raid10 pool, or even raid5, if you have really good backups.

Alright, thank you for the feedback.

 

I have a scheduled maintenance in a few days, I'll do that then.

 

I'll consult here to double check my plans.  

Back on the topic.

I am moving all my files from local_data to cache with Midnight Commander.

 

I would like to be sure of the next steps. I thinking:

1. scrub cache to start with a clean slate
2. stop array
3. remove drives from local_data
4. start array
5. stop array
6. assign freed drives to cache
7. start array
8. wait for the end of automatic raid1 balance
9. convert/balance for raid 10

Are there mistakes, missing or useless steps ?

Edited October 31, 20205 yr by ChatNoir
correct terrible english :-/

7 minutes ago, ChatNoir said:

• scrub cache to start with a clean slate
• stop array
• remove drives from local_data
• start array
• stop array

Scrub is only done on the data/metadata, if you're moving the data not much point in scrubbing after, you could (and should) wipe them completely with blkdiscard before adding them to the other pool.

I am not great at linux commands, is it simply a blkdiscard sdX or should I use options ?

blkdiscard /dev/sdX

 

P.S.: -beta30 requires -f to force if there's an existing filesystem.

Still on beta25. So not sure if it is necessary but I wanted to use it just in case.

The -f option is not recognized (or I am doing something wrong).

root@mediahub:~# blkdiscard -f /dev/sdf
blkdiscard: unrecognized option '--f'
Try 'blkdiscard --help' for more information.

 

I only have those options available:

root@mediahub:~# blkdiscard --help

Usage:
 blkdiscard [options] <device>

Discard the content of sectors on a device.

Options:
 -o, --offset <num>  offset in bytes to discard from
 -l, --length <num>  length of bytes to discard from the offset
 -p, --step <num>    size of the discard iterations within the offset
 -s, --secure        perform secure discard
 -z, --zeroout       zero-fill rather than discard
 -v, --verbose       print aligned length and offset

 -h, --help          display this help
 -V, --version       display version

For more details see blkdiscard(8).

 

Anyhow, the regular blkdiscard seems to have removed the partitions on the drives.

 

The RAID10 balance / conversion is finished.

 

Thanks @JorgeB !

20 hours ago, ChatNoir said:

The -f option is not recognized

Yep, that's only for -beta30 (and future releases).