2

I'm creating a NAS and I've been doing a lot of research on the different SSD caches. I have a 120gb ssd which is more than enough for the OS so I want to use most of it as a cache for the md raid5 array. Before I commit to this setup I want to know what happens if the ssd cache were to fail? I don't want to spend extra money to mirror the ssd cache.

I also have some old 30gb ocz vertex (from an RMA a long time ago but never reused) which are a piece of crap but could be used as a cache if bcache is ok with ssd caches dropping out.

If bcache is unable to do this are their other ssd cache options that would be fine? I've read that ZFS is okay with cache drives failing and continues as normal off just the drives but it seems to not be persistent and raidz is not as flexible as md raid (changing raid types and growing the raid). I guess I could use ZFS on top of md raid but it seems like that's redundant and I've had a lot of years of good experiences using XFS/md raid setup in the past (good performance and no issues through countless power failures).

tl;dr Building a NAS, want to use SSD cache, worried about effects of ssd failure

cowmix
  • 71

1 Answers1

1

According to kernel.orgs documentation and my research, by default, it's only using the bcache as a reading cache. So writing will not be faster, as it always writes to the slower magnetic discs.
So as there is only a copy of often-read data, you won't loose any data, if you remove the ssd.

If you enable writeback:

echo writeback > /sys/block/bcache0/cache_mode

you will lose data, if the ssd fails, as it firstly writes the data to the fast ssd, and then move it to the harddrives.

As always - a backup is neccessary if the data is important to you.

Community
  • 1
davidbaumann
  • 1,915