Longevity of your Nextcloud server - HDD - SSD -

After a few exchange in this message thread:
Now What? Long term safeguards on a fully completed setup - #11 by stratege1401

… I was joking about sacrificing some chickens to the god of “neverhappenbutitcould” regarding backup, longevity and some other aspect of managing a nextcloud server.

Quoting myself in order to avoid some back in forth between threads here:

stratege1401

  • Some info about reliability/backup…my experience.

I’ve been using nextcloud since … 2017.

My personal server has always been based on Debian and a soft raid6 lvm.

The hardware most of the time is based on regular “gaming” hardware except Hdd who are Proversion of Seagate ironwolf for Nextcloud data and a SSD for OS.

My raid6 array grown from 6TB to 10TB (LVM)…

My HA backup is much more simpler, and have only a 12TB HDD, rsync every 6 hours, hosted by my brother ( we are happy FttH users with 10G boxes free.fr )

I did change my hardware twice since 2017.
First time 5 years ago, rebuilt next 13.

Rebuilt my OS once, rebuilding next 17.

And after 5 years, as I had to change the SSD after smart alert ( samsung 840 ). Decided also to update my server totaly. Rebuilding OS and Nextcloud to 23 two weeks ago.

My raid6 had 1 disk failure in 6 years.

We are 21 users on it … Both personal and professional use…

Everytime I had to work on the server, I had the possibility to go to maintenance, activate HA manually…(luck, I guess… My 2 chickens sacrifice…)

And has i have a 3 months limite for share links, I am not very concerned by them…

My HA server is 4 years old, the HDD is new…

I could have gone for S3 external backup, but long haul pricing is bad…

stratege1401

It was in a 5 years span, not 6.

Thank you for the backblaze stats.

When i first bought my HDD, i spent time looking at the figure before making a choose. But, it was merely impossible to find any of the “good” PIN through regular EU shops …

Had to guess which was the closest model respecting my pricing ratio. I end up choosing the Pro Ironwolf model from Seagate ( PIN ST4000NE0025 )

I bought my disks in 4 shots ( 2 disks each time ) to avoid having same batches. Still having a brand new tested HDD in spare (6 power-cycle, 4 hours spin… brand-new !!! ) .

The Ironwolf seagate were all model: ST4000NE0025 with different built date and firmware ( 5 disk had the same date/firmware, not from the same factory… 3 had same date but another firmware, also not from same factory. ). I did not tried to mess with firmware and used them “as it”…
MTBF for those drive are “1.200.000 Hours” with a Maximum rate of 300TB/year ! Witch i am absolutely below ( around 193TB ).

The failing disk had a " Power On Hours = 83593", almost 3495 spinning days.
I experienced multiples smart error:

  • Reallocated Sector Ct blown over 100 ( bad bad bad )
  • G-Sense Error Rate was reaching over 7000, the others disks were closed to 900.
  • ECC-Hardware had reach alarm value.

Hopefully, at that time, i was still eligible for the 5 years warranty and i exchanged the failing disk.

Today, i have 4 disks over 92000 power on hours, one around 7000 power on hours…
The warranty program will be reach between december 2022 and February 2023. Not hoping for any failures before the end of the program.

My next server update-rebuild will be in 2 to 4 years… At that time, i will buy a full array of new disks…

now

Some very interesting stats from backblaze about 2021 Q3 are available :

And you ?

I think it would be interesting to know what kind of hardware you are using as ssd/hdd and their usage…

  • Today, i am using SSD 870 QVO 1 To for OS/Boot on both servers.
    – Primary server is based on Debian/LVM2/RAID6

  • 10TB Raid6 array of 5 Pro Ironwolf model from Seagate ( PIN ST4000NE0025 4TO ) on my primary server.

  • 12TB, single disk HGST UltraStart DC/HC550 WUH721816ALE6L1
    – Secondary server is based on UNRAID

I have built my two nextcloud servers (HA) using mostly “gaming” hardware, except for the HDD.
Both server are based on AMD Ryzen 5 3400G with Radeon Vega Graphics, 8 cores.
They both have Corsair Rm750x…

The primary server is Debian based, with a soft LVM2 raid60 array of 10TB using the Pro Ironwolf model from Seagate ( PIN ST4000NE0025 ). I use a PCI sata controller to run my raid6 array ( chipset based on a Marvel 88SE9215 ).

Secondary run under UNRAID with a Debian11 VM

  • Today, i am using SSD 870 QVO 1 To for OS/Boot on both servers.
    – Primary server is based on Debian/LVM2/RAID6

  • 10TB Raid6 array of 5 Pro Ironwolf model from Seagate ( PIN ST4000NE0025 4TO ) on my primary server.

  • 12TB, single disk HGST UltraStart DC/HC550 WUH721816ALE6L1
    – Secondary server is based on UNRAID 6.9.2

What is the question? Just listing specs?

Hi,

not completely…

My idea was more like a three step question:

  • of course, a glimpse of your hardware, specifically for a nextcloud useage.
  • what drives you to choose this hardware…could be interresting to others users.
  • crash, glimpses, bugs and how you manage them …

… linked this the goal of having a strong reliable system. Nextcloud related.

Now, who care’s … :)…remind me of @JimmyKater DFTT !!! :wink:

I don’t know if the hardware itself makes such a huge difference. For a single disks you always have to consider failures, and at some point they will happen. With some luck, you can anticipate with SMART.

Hardware is important…

Last week, i build a new nextcloud server for a friend. The PCI sata extender was based an a ASM1064 chip. Impossible to manage the full 4 lines… I bought 3 different extender. All ASMbased ( 1064/1164/1166). With debian, only the 1166 worked fine. The two others were behaving in a strange way…