Re: sabnzbd in Docker in ubuntu vm on proxmox... destined to be slow?
Posted: August 23rd, 2023, 11:48 am
Did this get resolved? What was the solution?
I just about to add this to proxmox SABnzbd Forums
sabnzbd in Docker in ubuntu vm on proxmox... destined to be slow?
Page 2 of 2
Re: sabnzbd in Docker in ubuntu vm on proxmox... destined to be slow?
Posted: August 29th, 2023, 8:36 am
I just about to add this to proxmox Re: sabnzbd in Docker in ubuntu vm on proxmox... destined to be slow?
Posted: August 29th, 2023, 12:26 pm
I should update, I did get my speed issues resolved.
For details, config is as follows.
Sabnzb installed in a lxc container in proxmox (5800x, 64gb)
Truenas installed as a vm with sata controller pass through (for direct disk access)
Two disks defined - 1) ssd for temp/scratch use, and 2) rust for media storage - seagate 4tb 2.5" SMR 5400rpm POS
Above 2 disks shared via NFS4 to proxmox
direct unpack enabled
Disks mapped to container using bind mounts.
There are several places where sync can be defined - more info https://www.avidandrew.com/understandin ... ching.html
1) zfs dataset
2) client*
For #1, the initial setting was "standard" which means allow client to define sync status
I tested 2 variations for client - sync and async.
With sync, speeds were abysmal, 15-30 MB/s on a disk capable of 130MB/s
With async, speeds improved but jumped beween 60-100 MB/s.
Setting zfs dataset sync to "DISABLED", and leaving client at async resulted in the best speeds, of 120-130 MB/s consistently (per iostat and zpool iostat 1). While this is not the safest configuration (in case of powerloss), it doesn't really matter for this data type (media).
For details, config is as follows.
Sabnzb installed in a lxc container in proxmox (5800x, 64gb)
Truenas installed as a vm with sata controller pass through (for direct disk access)
Two disks defined - 1) ssd for temp/scratch use, and 2) rust for media storage - seagate 4tb 2.5" SMR 5400rpm POS
Above 2 disks shared via NFS4 to proxmox
direct unpack enabled
Disks mapped to container using bind mounts.
There are several places where sync can be defined - more info https://www.avidandrew.com/understandin ... ching.html
1) zfs dataset
2) client*
For #1, the initial setting was "standard" which means allow client to define sync status
I tested 2 variations for client - sync and async.
With sync, speeds were abysmal, 15-30 MB/s on a disk capable of 130MB/s
With async, speeds improved but jumped beween 60-100 MB/s.
Setting zfs dataset sync to "DISABLED", and leaving client at async resulted in the best speeds, of 120-130 MB/s consistently (per iostat and zpool iostat 1). While this is not the safest configuration (in case of powerloss), it doesn't really matter for this data type (media).