Search results

I have a spare free SAS slot in my server. I insert SATA and SAS drives for single point backups instead of using USB.

You should absolutely avoid USB for any serious storage :)

I was more thinking about out-of-order writes caused by disk firmware or USB. That would make ZFS quite unsafe wrt sudden disconnects.

It really isn't quite that clear cut. Reorderings from hardware (or rather, the firmware controlling that hardware) can happen and would leave a ZFS in an inconsistent state on a random disconnect. Even when ZFS only ever issued write sequences that as originally ordered were incorruptible. USB...

I hesitate about this. ZFS can be very unamused on a sudden USB cable connection problem and the like. Keep in mind there is no fsck for ZFS.

I wonder how much the missing cable would run you. Tens of thousands I guess, or maybe even 100k. Depends on whether tranceivers are also missing.

Are these just networked dual-CPU nodes, or do they have some system to share memory between cores across nodes?

That does not work when the ZFS pool has features active that the new OS does not (yet) support.

I run a BR writer on a SAS2008 right now.

I have a 4-processor G34 board using 1U heatsinks with random fans blowing at them. It's not thermically stable in summer, aka about now. It isn't very valuable, so I would prefer not to pay 4x $60 on new HSFs. Best I see are the Supermicro HSFs, but you have to pay extra for the G34 clamps...

Are you sure that you aren't better off with one powerful generic graphics card? You will lose some RAM, but as you say you don't yet know how much RAM you need with the GPU/accelerator. Either way I don't think you have to worry about PCIe bottlenecks. If you are in the US this system with 4...

Nice build. Optane FTW.

What mrpasc said. If the data coming out of that machine have actual function ECC is much better.

I am not convinced. I would expect that the same problem could appear on your new machine.

How do you know that?

Seems like a good buy when you want - raidz3 - and hence 8 drives total - and your budget can't cover 8 large drives plus spare(s)

Spam in 3... 2... 1...

That is all true. My point, however, is that with NFS this part of the problem isn't made worse compared to a VM-local snapshot.

Well :) https://github.com/freebsd/freebsd-src/tree/main/sys/fs/nfsclient Just joking. It should just be obvious. The server has an open view of all the files it shares to the NFS client. You could chroot into the tree if you wanted. As a result there is no buffering that would be affected by...

Yeah, most Unixens support that. I run PXE+NFS for most of my machine zoo for many years. The guest OS handles NFS completely different from the situation where it has a local block device for a non-networked filesystem. It is always consistent against the server. Not is why you can now safely...