If this is your first visit, be sure to
check out the FAQ by clicking the
link above. You may have to register
before you can post: click the register link above to proceed. To start viewing messages,
select the forum that you want to visit from the selection below.
Phoronix, Linux Hardware Reviews, Linux hardware benchmarks, Linux server benchmarks, Linux benchmarking, Desktop Linux, Linux performance, Open Source graphics, Linux How To, Ubuntu benchmarks, Ubuntu hardware, Phoronix Test Suite
Working in HPC, I was happy to see Lustre being merged into staging 5 years ago, for all the usual reasons out of tree drivers are a pain.
But at this point, I can't really blame gregkh. It always was a code dump thrown over the wall, while development continued in the out of tree version, with no progress towards a upstream first approach. All the users kept on using the out of tree version as that had all the features and bug fixes.
That being said, there were a bunch of (idealistic?) developers working on the staging version, trying to massage it into shape for mainline. It would be a huge shame if all this work is wasted. In the end, it wasn't enough, and the upstream version diverged more and more.
I do hope to see Lustre in mainline at some point, but clearly this would need commitment from the core developers, and not just be an idealistic project by some minor players.
Working in HPC, I was happy to see Lustre being merged into staging 5 years ago, for all the usual reasons out of tree drivers are a pain.
But at this point, I can't really blame gregkh. It always was a code dump thrown over the wall, while development continued in the out of tree version, with no progress towards a upstream first approach. All the users kept on using the out of tree version as that had all the features and bug fixes.
That being said, there were a bunch of (idealistic?) developers working on the staging version, trying to massage it into shape for mainline. It would be a huge shame if all this work is wasted. In the end, it wasn't enough, and the upstream version diverged more and more.
I do hope to see Lustre in mainline at some point, but clearly this would need commitment from the core developers, and not just be an idealistic project by some minor players.
Exactly this + I have no clue how bad the code for lustre actually is. Working in physics and on multiple clusters and we have severe lustre issues on a monthly basis with roughly one major file damaging incident per year per cluster. And I am not counting the "Oh, suddenly lustre is in read only mode until someone sacrifices a goat" or the silent file degradation issues. This also does not seem to be caused by the underlying hardware as it happens on separate clusters with different hardware and buy dates.
Is Lustre something which is built into a kernel module? Kernel modules would be natural for a filesystem. This makes it somewhat less of a problem to have it non-mainline, since you don't need messy kernel patches that conflict with each other. Kernel patching can be one of the worst things to deal with.
Comment