That's a great test environment and it really helps drive away some of the misconceptions about ZFS.
It's a lovely filesystem and it can work in very stodgy ways with zetabytes on Serious Hardware, but it can also work with relatively small devices on silly hardware.
Another example: I've got a ZFS scratch disk that has seen dozens of terabytes written. It's an old ~120GB Intel-branded SATA SSD that is connected with a very cheap SATA to USB3 adapter cable. I've been using it this way for years.
All of the Internet lore has told me that this must result in resolute failure. Whether because USB is broadly not a serious data bus, or because ZFS doesn't like non-serious things, or because 1 copy is 0 copies, or because SSDs are fragile creatures that hate write cycles, or because I've created a portal to hell, or something else: The lore states that it must result in failure, and usually stipulates that ZFS would amplify this concern.
But it's been great. Really. Zero issues -- except when I unplugged it on accident while it was hot...but even then, it came back fine. (Of course it did. ZFS is intended to be robust in the face of unplanned failures regardless of cause.)
And because ZFS: If it ever has any issues, then they'll be reported and logged either in the normal course of doing things, or during a periodic scrub. Every read is checksummed (even in normal operation), and every discrepancy is reported. It's been clean.
It is my intent to keep it as a scratch disk until it becomes problematic, and my further hope is to be able to analyze those problems once they occur.
It won't last forever, but it's lasted for quite a long time so far. (A little bit of inquiry with smartctl and some math just now shows that it has endured writes equivalent to ~43,000% of its capacity.)
It's a lovely filesystem and it can work in very stodgy ways with zetabytes on Serious Hardware, but it can also work with relatively small devices on silly hardware.
Another example: I've got a ZFS scratch disk that has seen dozens of terabytes written. It's an old ~120GB Intel-branded SATA SSD that is connected with a very cheap SATA to USB3 adapter cable. I've been using it this way for years.
All of the Internet lore has told me that this must result in resolute failure. Whether because USB is broadly not a serious data bus, or because ZFS doesn't like non-serious things, or because 1 copy is 0 copies, or because SSDs are fragile creatures that hate write cycles, or because I've created a portal to hell, or something else: The lore states that it must result in failure, and usually stipulates that ZFS would amplify this concern.
But it's been great. Really. Zero issues -- except when I unplugged it on accident while it was hot...but even then, it came back fine. (Of course it did. ZFS is intended to be robust in the face of unplanned failures regardless of cause.)
And because ZFS: If it ever has any issues, then they'll be reported and logged either in the normal course of doing things, or during a periodic scrub. Every read is checksummed (even in normal operation), and every discrepancy is reported. It's been clean.
It is my intent to keep it as a scratch disk until it becomes problematic, and my further hope is to be able to analyze those problems once they occur.
It won't last forever, but it's lasted for quite a long time so far. (A little bit of inquiry with smartctl and some math just now shows that it has endured writes equivalent to ~43,000% of its capacity.)