6 ms·
I’m not a filesystem admin, but we at LLNL use OpenZFS as the storage layer for all of our Lustre file systems in production, including using raid-z for resilie
by trws 3y ago
I’m not a filesystem admin, but we at LLNL use OpenZFS as the storage layer for all of our Lustre file systems in production, including using raid-z for resilience in each pool (order 100 disks each), and have for most of a decade. That combined with improvements in Lustre have taken the rate of data loss or need to clear large scale shared file systems down to nearly zero. There’s a reason we spend as many engineer hours as we do maintaining it, it’s worth it.
LLNL openzfs project: https://computing.llnl.gov/projects/openzfs https://computing.llnl.gov/projects/openzfs
Old presentation from intel with info on what was one of our bigger deployments in 2016 (~50pb): https://www.intel.com/content/dam/www/public/us/en/documents/presentation/benefits-zfs-for-lustre-deployments.pdf https://www.intel.com/content/dam/www/public/us/en/documents...
- muxator 3y agoIf I'm not mistaken the linux port of ZFS that later became OpenZFS started at LLNL and was a port from FreeBSD (it may have been release ~9). I believe it was called ZFS On Linux or something like that. Nice how things have evolved: from FreeBSD to linux and back. In my mind this has always been a very inspiring example of a public institution working for the public good.
- rincebrain 3y agoFreeBSD had its own ZFS port. ZoL, if my ancient memory serves, was at LLNL, not based on the FreeBSD port (if you go _very_ far back in the commit history you can see Brian rebasing against OpenSolaris revisions), but like 2 or 3 different orgs originally announced Linux ports at the same time and then all pooled together, since originally only one of the three was going to have a POSIX layer (the other two didn't need a working POSIX filesystem layer). (I'm not actually sure how much came of this collaboration, I just remember being very amused when within the span of a week or two, three different orgs announced ports, looked at each other, and went "...wait.") Then for a while people developed on either the FreeBSD port, the illumos fork called OpenZFS, or the Linux port, but because (among other reasons) a bunch of development kept happening on the Linux port, it became the defacto upstream and got renamed "OpenZFS", and then FreeBSD more or less got a fresh port from the OpenZFS codebase that is now what it's based on. The macOS port got a fresh sync against that codebase recently and is slowly trying to merge in, and then from there, ???