SSD lifetime in software RAID 1

Question

We're looking at use two SSD in a software RAID 1 raid on CentOS. We have tested performances and it is the same as a SSD alone.

But we fear that because of mdadm, lifetime of the SSD will shorten. Here is the warning of Redhat team :

Red Hat also warns that software RAID levels 1, 4, 5, and 6 are not recommended for use on SSDs. During the initialization stage of these RAID levels, some RAID management utilities (such as mdadm) write to all of the blocks on the storage device to ensure that checksums operate properly. This will cause the performance of the SSD to degrade quickly.

If mdadm write on all the disk on initial check seems OK, but how does it check thereafter ?

Answer 1

Theoretically, as long as you don't rebuild the array, the initialization will not happen again, so no more extra ware.

<mumbling> However, writing to all the blocks in one go can easily overwhelm the disk controller and cause major problems. Normally SSDs keep a pool of pre-erased blocks to allow fast write access. When that pool is depleted, the controller will have to erase blocks on-the-fly to write into them. This takes significantly more time so when these requests pile up, the earlier ones will start to timeout. The software may mistake this for a bad block or crash, in which case you may have to start again. Obviously this is not good, but hardware and software are both improving, so this should be less and less significant. </mumbling>

I don't see why can't you use it with mdadm and any reason for siginificant flash life loss if you're careful. It might be a good idea to pratice on some hard disks and figure out how to configure all the parameters, e.g. chunk size.