>Thanks Al,
>
>Very informative and now a little troubling to me since I just replaced my dual 2TB Seagate HDDs and they appear to be SMR and not CMR. I have them in my RAID 1 configuration on my desktop machine.
>
>It's only been 1 week since the switchover. One of the previous Seagates had failed and the RAID was degraded.
>
>What should I be looking for in terms of trouble?

I haven't followed that story in depth. From what I've seen, some people could not rebuild ("re-silver") mirrored arrays because the newly inserted SMR drive would pause for long periods during writes, which the RAID controller would interpret as a failure. Rebuilding an array is one of the most write-intensive operations. If you've already done it successfully, you're at least past that hurdle.

OTOH - if you replaced both drives, maybe you imaged the single drive, replaced it with the 2 new drives, then restored from the image (?) That would have avoided the situation where you have an array with mixed SMR and CMR drives. Doing that might be quicker than replacing the failed/missing drive with an SMR, resilvering, then replacing the old drive with the other SMR and resilvering again.

Going forward you might see slower write performance than with CMR. However if you've enabled write cache in Windows you may not notice it.

Looking a little deeper, the reports I'm seeing of resilver failures are on Linux using ZFS. Linux filesystems generally recommend software RAID, not enterprise hardware RAID or the so-called FakeRAID (i.e. Intel RST) you typically get on desktop mobos. Here's a link: https://www.ixsystems.com/community/threads/list-of-known-smr-drives.83993/

It looks like there are multiple types of SMR drives, which may or may not work properly depending on the OS. The outrage seems to be that makers are slipstreaming these drives into their sales channels without any warning, so caveat emptor.

I would definitely avoid mixing SMR and CMR drives in the same array. It could also be argued that if SMR has poor write performance, it should not be used in RAID at all, because even with CMR, rebuild times with large capacity drives are already dangerously long.

For lurkers, why are long rebuild times dangerous? Simple redundant RAID types can tolerate a single drive failure without data loss (but with degraded performance). For example, a RAID5 array of 5 drives can lose a single drive. The idea is, when a drive fails you replace it with another of the same capacity or larger, and the controller will automatically rebuild the array and restore redundancy.

However, until the rebuild process is complete, there is no redundancy. Loss of another drive means complete loss of the array and all its data. This threat is not insignificant; if the original 5 drives were all the same make/model and from the same manufacturing batch, and one has failed, there's a significant chance another could fail in the very near future. In effect, while the array is still degraded (i.e. rebuild not complete) there's danger of losing the array to another failure. Long rebuild times means a long time in the danger zone.

There are several ways to combat this issue:

- Use a newer RAID level which can tolerate 2 or more drive failures
- Install and configure a hot spare. It will be used as the rebuild target as soon as a failure is detected, minimizing the degraded time (no need to source, order, receive and install the replacement drive)
- Installing a cold spare is the same idea except the drive requires manual intervention to be used as a rebuild target

And of course, everyone should have good backups :) "There are 2 types of people - those who are true believers in backup, and those who are about to become true believers in backup".

Update: here's another link with a pretty good summary of the current SMR drive situation: https://www.tomshardware.com/news/sneaky-marketing-toshiba-seagate-wd-smr-drives-without-disclosure