There's a problem or a basic conflict between sysutils/zxfer and sysutils/zfstools --- zfstools will happily make and delete snapshots on a schedule. zxfer seems to crash if, after it scans the pool, a snapshot gets deleted.
It would be better if it re-read the pool configuration and then continued to work.
I am looking at making it re-fetch the list of snapshots if the list it has is more than a minute or two old, but have not gotten around to implementing it.
I am planning a totally new tool to replace zxfer, because it fundamentally lacks the ability to handle things like resumable send and bookmarks.
Anyway, please file this upstream with the port here: https://github.com/allanjude/zxfer/issues/new
Assign Allan his own PR.
(Maybe it's a good idea to change MAINTAINER to allanjude@FreeBSD.org on all ports maintained by firstname.lastname@example.org ?)
correction: PR was opened by dgilbert but it's still better to assign it to FreeBSD user name
Any news here?
This is still a known issue.
For people coming to check the status of this issue:
This comment (2020) was quite important for my use-case:
> You should not create snapshots on the destination of the zfs replication.
(In reply to Sam from comment #6)
Yeah... you're not reading this correctly. A good example is that you're running (say) sysutils/zfstools provided zfs-auto-snapshot. Then you run a zxfer. While zxfer is running, zfs-auto-snapshot runs on the source machine.
zxfer only looks for the synchronization list once when it starts. zxfer also fails if a snapshot is deleted before it gets around to it.