Best way to replicate entire pool to external hard drive

TrueNAS General
1

Hi,

I have one pool (tank) with multiple datasets, each with multiple children.

I have an external hdd dock connected to the TrueNAS server via usb3. Works well and I can access it like an internal drive to the server.

I want to create a full backup of tank including absolutely all data, snapshots, everything. I want a complete replica so I can restore the Truenas system in the case of disaster (a backup).

I will run the replication task manually - connect a drive, import the pool, run the replication task.

I want it to mirror the snapshots completely, so as
snapshots are deleted on Tank according to retention policies I want them to be removed from the external drive when running the replication. As new snapshots are added to Tank, I want the replication to mirror these across.

I canā€™t seem to achieve this.

I have tried selecting Tank as the source and the ā€˜recursiveā€™ option and it indeed replicates all the data and snap shots. But the next time I run it, it does not clean up any deleted snapshots and the snapshot number on the external drive just grows and is no longer aligned with Tank.

I have tried using the ā€˜full replicationā€™ option, but then the ā€˜recursiveā€™ option is not available any more. I can then only select the child datasets of tank, which means all the auto snapshots of app and system data donā€™t seem to then replicate (which are located at Tank/app etc) as it throws an error if I select both child and parent datasets, ie. you canā€™t select both as data overlaps I am guessing?

What is the easiest way in this setup to keep a full copy of my Truenas data as a backup (absolutely everything) that remains a full mirror of the original pool, so I can recover everything in its entirety?

I must be missing something here.

BTW, I have overlapping snapshots set up for each of my data sets, where I have set it to recursively snapshot child datasets as well as the parent. I have also set retention periods for each. I have set the replication option of ā€˜Same as Sourceā€™ for retention period.

How do I mirror my Truenas hard drive to an external drive using the gui, preferably in an incremental way to save time and data transferred (changes only). I mean could dd the disk but seems dumb given power of zfs.

Any help much appreciated!

2

When i set up the local replication task, these are the settings i use:

  1. Tick ā€œrecursiveā€ = on
  2. Tick ā€œinclude dataset propertiesā€ = on
  3. ā€œDestination dataset read only policyā€ = set
  4. ā€œSnapshot retention policyā€ = ā€œsame as sourceā€

If this is your regular backup Iā€™d also tick ā€œsave pending snapshotsā€ so the external destination can always be brought up to date.

1 Like
3

Thanks very much for the advice, I am playing with your suggested settings now.

I am finding that when I select the pool (Tank) and use ā€˜recursiveā€™ it says 1 or 1 snapshots will be replicated. If I select the pool and also each of the next level down child datasets (this children have children but I am not selecting those), it gives me a much higher number of snapshots which will be replicated. Does selecting just the pool not drill down to the entire hierarchy of datasets even when ā€˜recursiveā€™ is selected? It looks like you also need to select each of the top level child datasets to get recursive to work properly?

Do you use naming schema or regular expressions? If I use naming schema and donā€™t specify any schemas in ā€˜also includeā€¦ā€™ I donā€™t get many snapshots replicated. If I add all the naming schemas to cover all snapshots, I get more but still only about half of those on the system. If I use regular expressions and .* I get 526 of 526 snapshots selected (which is much better!), but the system actually has 600 snapshots on it. I donā€™t know why it is not pulling everything!

It is all quite confusing - I just want a complete mirror image of the TrueNAS pool sent to an external disk.

Thanks for any further help.

4

I do not select the whole pool. Instead, I tick off the individual top-level datasets that I want replicated. Selecting a whole pool didnā€™t work for me.

I use the regular automatic naming that the TrueNAS develops for snapshots.

I would consider selecting the tickbox for a complete replication from scratch until you get the result you were looking for.

5

Thanks for that advice. Selecting the pool, with recursive, seemed to fail create my Datasets, but selecting all the Datasets did.

It was annoying that they all go a non-existent userā€™s ID, because WTF annoyingly, User IDs can no longer be specified at user creation or later, e.g. via shell commands, like in normal Linux/FreeBSD. So, Iā€™ll have to waste time redoing all the Dataset permissions! What were Ix thinking?