Old Parts or New Build

[Stux]

I fully agree with those statements but in my mind the RAIDZ3 is addressing a convenience issue because restoring that much data is a pain in the rear and can take considerable time. Otherwise we would all just used striped, no RAIDZ, and simply restore our backups after a failure, right :D

Well, no. I have a failed drive, I replace it, it takes 2-3 days to resilver. I don’t freak out as I have RaidZ2. And a backup. And I don’t intend to ever have to go to the backup, so the amount of time it takes to restore from backup is irrelevant, but I do have one.

ZFS has resilvering optimizations coming down the line too

I’m not convinced it’s time to start recommending raidz3 by default, just because it can take a few days to resilver.

 

[joeschmuck]

I’m not convinced it’s time to start recommending raidz3 by default, just because it can take a few days to resilver.

That is okay, we have a difference of opinion but I'd rather point this out to someone building a system and who has plans to store a ton of data vice letting them find out later that it was an option that they may have desired due to slow resilvering.

ZFS has resilvering optimization’s coming down the line too

I've heard and honestly I can't wait for it to materialize in a stable version of FreeNAS.

 

[Chris Moore]

I’m not convinced it’s time to start recommending raidz3 by default, just because it can take a few days to resilver.

I have a server at work using 6TB drives that takes about 3.5 days to resilver a drive. Too much data. My home system resilvers a drive in around 4 hours. Much less stressful. In both situations, I have RAIDz2 and cold spare drives ready to go when the need arises.

That is okay, we have a difference of opinion but I'd rather point this out to someone building a system

It is good to know the options up front, but I don't think there is just one answer, it depends on your tolerance of risk and the convenience factor. For example, I have a server at work configured with 10 RAIDz2 vdevs and the pool also has 5 hot spares. I wouldn't do hot spares in a home server, and I am not sure I really want them in this server at work, but it was like that when I arrived here. Maybe it was a good thing because the system had not been maintained in a while and the first time I checked it all 5 hot spares were in use because of drive failures that had happened while nobody was looking.

 

[joeschmuck]

As long as we don't force our own beliefs on someone and provide them with the options and pros and cons. I'm not sure I'd run a RAIDZ3 either but I'd seriously consider it if I were to need that much data and didn't want to worry about the time it takes to restore the data.

the first time I checked it all 5 hot spares were in use because of drive failures that had happened while nobody was looking.

Bet that was both a good feeling and scary one as well.

 

[Chris Moore]

Bet that was both a good feeling and scary one as well.

You bet it was scary. I guess I didn't give enough details to show just how scary. Not only were all the hot spares in use, in one of the vdevs, both of the 'redundancy' drives had failed, so in that vdev, if I lost a single additional drive, the whole pool would have been gone. That system is quite slow (by my way of thinking) and it took more than a day for each drive to get the pool fully healthy again.
I wasn't quite as nervous after replacing the first couple drives but it was still more than a week where I was watching and waiting for something else to go wrong.

PS. It used all the cold spares we had, for the size this server uses, and I needed to borrow a couple drives from another department, then order more to pay them back and have some additional cold spares of our own.