ZFS in 12 beta - is it considered solid yet against data loss?

[Stilez]

I've been looking over the Jira posts against v12/OpenZFS/Samba/iSCSI, these being the main functions I use. The main issues affecting ZFS seem to be either GUI/middleware, or failure of ancillary functions like send/receive or device hotswap. Not issues that actually damage the pool or compromise data?

But v12 is essentially a pivot of ZFS to an entire new codebase (ZoL version). Thats a huge step, even if the code is 98% identical. So I would like to check - how safe is the current 12 beta considered by "those in the know", against pool data loss/damage. Have we seen any instances of pool damage that weren't attributable to.usual causes such as inadequate hardware or insufficient redundancy? Are send/recv from CLI, and Samba/iSCSI, basically reliable?

I use ZFS CLI commands to manage my pool, so I'm less worried about pure GUI issues and few of the jira issues remaining seem relevant to.my core needs.

Tl;dr as we move towards 12 release, how safe is my pool.itself to migrate to v12/openzfs 2.0, even if not all higher level TrueNAS functions are ready for.prime time? And if I use my normal pool management (ZFS CLI, and samba/iscsi), are we seeing any risk of data damage relevant to my uses?

 

[Kris Moore]

I've been looking over the Jira posts against v12/OpenZFS/Samba/iSCSI, these being the main functions I use. The main issues affecting ZFS seem to be either GUI/middleware, or failure of ancillary functions like send/receive or device hotswap. Not issues that actually damage the pool or compromise data?

But v12 is essentially a pivot of ZFS to an entire new codebase (ZoL version). Thats a huge step, even if the code is 98% identical. So I would like to check - how safe is the current 12 beta considered by "those in the know", against pool data loss/damage. Have we seen any instances of pool damage that weren't attributable to.usual causes such as inadequate hardware or insufficient redundancy? Are send/recv from CLI, and Samba/iSCSI, basically reliable?

I use ZFS CLI commands to manage my pool, so I'm less worried about pure GUI issues and few of the jira issues remaining seem relevant to.my core needs.

Tl;dr as we move towards 12 release, how safe is my pool.itself to migrate to v12/openzfs 2.0, even if not all higher level TrueNAS functions are ready for.prime time? And if I use my normal pool management (ZFS CLI, and samba/iscsi), are we seeing any risk of data damage relevant to my uses?

The pool stability seems real good at this time. We've already seen nearly 1000 systems upgrade to it and have had no reported issues of pool corruption, data-loss, etc. The handful of issues we're investigating all are related to performance, but mostly at the much higher end of things that we've only seen in our performance labs.

 

[Stilez]

@Kris Moore
I'm a high performance obsessive ;-) but right now my pool is coming towards the end of many years of suffering because of lack of allocation classes/devices, or any other metadata preload/efficiency mitigation that's either in OpenZFS 2.0 or actively being worked on.

(My pool is about 4.5x dedupable with huge VM files on 140 TB of SAS3 HDD, so its been a case of either dedup or huge extra HDDs, regardless of stuffing optane L2ARC and hundreds of GB ECC ram onboard for the DDT... and its awesome to finally see dedup preload, warm.L2ARC, and other DDT/Metadata IO mitigations coming down the line! Thank you for your and ixsystems part in this while I think of it!!!)

So I can't wait to move to OpenZFS 2.0. Even if cutting edge performance isn't spot on, the performance hit of 4k DDT IO I currently have is huge so its a big net win for me ...... provided early adopting the beta doesn't lose data, or fidelity of send/recv, or usability via Samba.

And that's really my question. If all I care about is ZFS and send/recv not trashing the pool or junking my backups, and Samba working for file access, how would 12 beta stand even though I know its not even production yet?

 

[jasonsansone]

@Kris Moore
And that's really my question. If all I care about is ZFS and send/recv not trashing the pool or junking my backups, and Samba working for file access, how would 12 beta stand even though I know its not even production yet?

I haven't upgraded any main systems (and won't until release), but I'm running all backups on the BETA-1. I haven't noticed any issues at all. SMB, jails, VM's, and replication are working fine. Take it for what its worth.