How having an online backup saved my home server today
I have a physical server running at home that has a few services running on it, including Nginx Proxy Manager to route external network traffic to an OpenWebRx SDR that I host at home. The server boots off its own NVMe internal drive, but has two 4 TB 2.5″ drives connected by USB to it, on which the Docker container data sits. The server runs Open Media Vault OS which monitors everything.
The backups are configured with Timeshift backing up the OS boot drive to external drive 1. My desktop computer backs up over the network also to external drive 1. At about 02:00 or so, there is a rsync that runs to mirror external drive 1 to external drive 2. So external drive 2 is a full backup of everything.
I noted about a week ago I was getting SMART reports mailed to me complaining that external drive 1 had picked up about 32 uncorrectable errors. I checked its age and yes the drive has been running continuously for over 5 years now. These are Seagate Barracuda drives, so actually are not designed for this type of punishment. They are anyway due for replacement (with Seagate IronWolf drives, which are better for NAS type operation).
I also decided to order the SATA port kit to allow me to connect the new drives via eSATA connections. This would allow better access from the BIOS for managing the drives, and also some faster performance than USB. This kit is on order from the Tariff War country and still on its way.
Just yesterday, I decided to order the first Seagate IronWolf drive from a local supplier. I ordered it with normal “cheaper” courier delivery. But little did I know that the first external drive was about to go downhill. I noticed yesterday’s SMART report now mentioned 80 unrecoverable sectors. Ouch, this means the situation is actually worsening. But there are millions of sectors, so surely this will hang on OK for a few more days… External drive 2 was still OK as, although it was spinning for 5 years, it only did a rsync every 24 hours.
Well, no, it seemed it would not. At about midnight last night, I was just going to check something on Jellyfin on my phone, before I went to sleep, and I was prompted for the login password. The password was not accepted, and I struggled a bit further until I noticed most of my Docker services were not responding. I rebooted the server but still nothing. So, I stumbled through to my study at 01:00 and switched the desktop PC on. After logging into OMV I noticed external drive 1 was not showing.
Long story short(ish), I realised it had died. I did manage to remove that drive from the fstab boot config, and to set external drive 2 to use drive 1’s description, etc. OMV then was able to recognise drive 2 as drive 1, and everything worked again. So, it is all up and running, but without any backup drive now.
There was quite a bit of fiddling needed to remove drive 2’s references, as OMV kept complaining that the drive config could not be removed as it was still referenced somewhere.
So another lesson to try to make this easier in future, is now that I will have the eSATA connections to the drives (OMV does not want to do RAID with USB connected drives), I will rather set up a RAID 1 mirroring. This is not so heavy on the drives, as it just does a write to drive 2 every time anything gets written to drive 1. But there is no heavy-duty RAID rebuilding taking place. RAID 1 will work pretty well, much like my rsync did, but will be more real-time, so I won’t lose a few hours of changes if there is a crash in future.
Also, any RAID does mean you have to format the drives and use the accepted RAID naming hierarchy. But something interesting I picked up for RAID 1 mirroring, is that you can configure it using just drive 1, and tell it drive 2 is missing for now. Then you can copy your data over from drive 2 to drive 1, configure OMVs paths for the containers to see new locations, and then format and mirror to drive 2.
So why did I not go for a full external NAS or a DAS (Direct-Attached Storage) without NAS functionality)? Well, if you want to properly monitor and manage your drives via SMART data, a DAS is not really an option if RAID is used, as OMV will not see the individual drives. Also, my Lenovo server does not have port multiplying functionality, which is needed to recognise multiple SATA drives across a single eSATA link.
A full NAS of course is probably ideal as all the CPU and bandwidth sits inside the NAS, as well as the management software. But a 4-bay NAS is not really affordable for me.
So after looking around for quite a while I actually found an interesting drive enclosure that will house both 2.5″ (my current drives) as well as 3.5″ drives (the new Seagate IronWolf, with the 123V external power needed for the 3.5″ drive) and it can connect via USB or eSATA. It is sort of the best of all worlds for transitioning between drives and setups. I’ll only be able to get another IronWolf drive next month, so this will work for now. It is the Vantec NexStar JX NST-358SU3-BK.
So, as soon as I have the drive enclosures and the new IronWolf drive has been tested, I’ll set up the new drive on its own in a RAID 1 configuration, copy the data from the remaining Barracuda drive onto it. Then let the Barracuda drive be drive 2 in the RAID config so that backups can continue. When the next IronWolf drive arrives, that can replace the Barracuda drive, and the RAID setup should rebuild it for backup use.
So quite a bit of drama, but it has made me realise two things:
* It is vital that you have full backups which are automated to run daily.
* RAID 1 is an easier way to mirror a drive, and makes any drive replacement a lot more seamless in future.
* Bonus lesson also is not to use desktop drives in a server type setup. For the price I paid for 2.5″ drives (powered by USB) I could have bought better suited NAS drives in 3.5″ form factor.
Yes, I suppose a proper NAS is going to be a better solution than software RAID in the longer run. So I will spend some time researching options around that later this year. The drives are not wasted as they can be used in a NAS as well now.
From
How having an online backup saved my home server today
I have a physical server running at home that has a few services running on it, including Nginx Proxy Manager to route external network traffic to an OpenWebRx SDR that I host at home. The server boots off its own NVMe internal drive, but has two 4 TB 2.5" drives connected by USB to it,...
#
technology #
storage #
backups