Now I have started many new threads and I’m sorry for that.
I got everything running in my lab VM so now I’m ready to take the step to take seafile in production.
I’ll have about 10 users, so it’s not so much but I do need the stability and max speed.
I have my server at a 1000Mbit/1000Mbit line.
The server runs on Windows Server 2016 and I’m using Hyper-V to have the seafile server on Ubuntu 16.04.
So, my quastion is, should I have one VHDX file with the ubuntu system on a SSD and then make a new VHDX file on a WD Red disk to store all data from Seafile?
Or is this just no necessary? Should I just have one big VHDX runnning on my WD red?
I would split the VHDX files into 250GB images each and then merge them together with an LVM within ubuntu as pure data disk.
It is better to put the OS on a separate disk, I think that 20-30GB should be fine, depending how much SWAP and TEMP space you need.
This sounds like a good solution to me. If you’d like to further optimize it and have some space left on the SSD you could also store the fs and commits folder from seafile-data/storage there. These folders are going to contain many, many small files. I’ll report from my installation running a few years with currently 800 GiB how much space that’d need.
root@home ./seafile-data/storage # du -sh commits/
985M commits/ # storage used
root@home ./seafile-data/storage # find commits/ -type f | wc -l
139899 # number of files (will be larger the more changes there are)
root@home ./seafile-data/storage # du -sh fs
22G fs # storage used
root@home ./seafile-data/storage # find fs/ -type f | wc -l
1770273 # number of files
Ok, then I’ll give the OS 30GB on my SSD, it 'll only run Seafile in ubuntu and nothing else so 30GB should bee good or do you think that I should have 50GB?
Then during the installation of Seafile I just choose to place the seafile-data folder on the other VHDX that are on my WD red system correct?
Since the storage/blocks (ty @shoeper) folder is going to hold most of the files, I’d put them on a drive that is managed through LVM to be able to increase this drive without the need to move any data and to be able to have the data not on one single massive VHDX file, but rather split on several.