So I get that unraid was built back when a 500gb hard drive was considered large. Thus a rebuild would take an hour or 2 if something when wrong and it was not a big deal to eject a disk from the array at the first signs of trouble.

 

In today's world of 14TB+ drives that take a full 24+ hours to rebuild, things are quite different. It is not a simple matter and it puts a lot of excess wear and tear on the drives.

 

A lot of use are forced to use old, low end, consumer hardware that are prone to issues like cables moving when swapping drives and causing a drive to drop etc.

 

In my case my server works fine but due to running a bunch of sata cables in a small space, anytime I opened the case I was almost assured to have some kind of cable issue in the coming week.

 

I have had to rebuild a drive around 8 times since I got unraid up and running a few months ago, all because of simple cable issues. Rebuilds alone have caused all my drives to use over half my rated yearly workload for the drives before the warranty is void. With normal use and some parity scans, almost all of them of out of warranty at this point.

 

So here is what I propose:

 

I would LOVE the ability to simply have unraid lock the array if it detects an issue with a disk(simply put it in read-only mode would be an easy way to lock it and prevent the parity from going out of sync). Let me know about the issue and give me a chance to fix it. In every case thus far all it needed was some cables jiggled and a reboot and all was well.

 

Having the choice to eject the drive and continue with an emulated drive is GREAT but I would really like to be given the choice. It would save so much stress, wear and tear along with downtime.

 

In this world of 24 hour rebuilds, ejecting a drive is a real pain as the server is basically down for a full day outside of your control.

I had this problem once too ages ago, since then I set my array not to auto start, that lets me check that all disks are there before i start the array and any drives are emulated. If there's something wrong with a disk i could start in maintenance mode which is essentially read only. It's saved me quite a few rebuilds since doing that.

 

I just had a case of parity sync issues and being able to reshuffle disks etc. and not stressing about a disk becoming emulated was one less thing to be concerned about.

Edited December 6, 20205 yr by Greg-Mega

6 minutes ago, Greg-Mega said:

I had this problem once too ages ago, since then I set my array not to auto start, that lets me check that all disks are there before i start the array and any drives are emulated. If there's something wrong with a disk i could start in maintenance mode which is essentially read only. It's saved me quite a few rebuilds since doing that.

 

I just had a case of parity sync issues and being able to reshuffle disks etc. and not stressing about a disk becoming emulated was one less thing to be concerned about.

I have auto-start disabled as well, the only way I know of to deal with a dropped disk is to use the new config option in tools and hope that everything is ok on the drive since the array would of kept writing to the emulated drive. This is what I am considering doing if it happens again but it is disconcerting to not know for sure that the data is ok.

 

I agree the paranoid stress of a drive dropping is starting really get to me. Right now I have a bay in my new hot swap cages that is questionable and I am scared to test it for fear of a drive dropping. Going to try a pre-clear in it I think to test it out.

 

Just in general I run around stupid scared that a drive will drop and force a rebuild on a perfectly good drive. When a drive actually is dropped I run around double scared that another drive might drop, it is not good for my health.

 

I went with unraid because it seems more stupid proof for less then ideal hardware, aka if a drive dies, I have parity but if that fails I only loose the data on that drive.

 

This though it putting more stress on the drives and me without the benefits afforded from BTRFS/ZFS.

 

I would not only be glad to simply have a write/read to an arry drive fail and it let me know there is an issue but prefer it for several reasons.

 

1: It lets me know right away there is an issue (email is great but I don't always check it while working) and exactly when the issue happened in a file transfer for example. I can then postpone the rest of the transfer for later.

2: If the drive is fine after a reboot, I can pick up right where I left off, downtime = >10 mins vs 24 hours to rebuild a perfectly fine drive.

3: If the drive is dead, I can chose to eject it from the array and emulate it / replace it.

 

Naturally there could still be the option to automatically eject a drive but I don't see what there can not also be an option to halt the array or make it read only when an error is detected. This plays much nicer with questionable hardware in a home server environment IMHO.

Trust me, I understand the stress of it (last 5 days solving parity issues not knowing if it was my data disks or parity disks at fault). 

 

Not really input for the feature resquest BUT - It sounds like if you resolved your hardware issues with the drives dropping - which they shouldn't - a lot of your stress would be eleivated (and also backups but I'm going to assume you have those as there's no substitute for backups!).

 

How would you prefer to be notified of issues? You could do something with IFTTT to turn unraids email notifications into some other sort of notification that you can see immediately.

15 minutes ago, Greg-Mega said:

Trust me, I understand the stress of it (last 5 days solving parity issues not knowing if it was my data disks or parity disks at fault). 

 

Not really input for the feature resquest BUT - It sounds like if you resolved your hardware issues with the drives dropping - which they shouldn't - a lot of your stress would be eleivated (and also backups but I'm going to assume you have those as there's no substitute for backups!).

 

How would you prefer to be notified of issues? You could do something with IFTTT to turn unraids email notifications into some other sort of notification that you can see immediately.

Yeah I want to fix the hardware issues, the fix is simple. Money lol Just too many drives crammed into too small of a case.

 

I really want to get a supermicro 4U chassis for future proofing but it seems the days of $200 server pulls are long gone.

 

That said I did upgrade this weekend to some rosewill hotswap bays that should help and getting an HBA tomorrow. Although over the long term there is still a high likelihood that an issue like this would pop up again.

 

Being a home server, I can't think of a situation where I would need it to eject and emulate a drive without user input.

 

If it could simply make the problem drive read only and allow the rest of the drives to work normally, that would be cool but even just putting the whole array into read only mode would suffice for 98% of home server use cases I can think of.

 

I would much prefer 10 mins to reboot then 24 hours to rebuild.

 

Far as notifications, the issue is more that I am not always around a device to notify me then an issue with the notification itself. Unlike a lot of people I don't live by my phone, heck I can go days without touching it more then moving it. Email is actually the best option for me, I just don't actively check it unless I am at my computer.

Edited December 6, 20205 yr by TexasUnraid