Forum: Author Hangout

The only reason I still use win10 is because of my Newsreader programme is windows based.

RSS feeds? Usenet groups? There are several excellent RSS readers for Linux. And if you use Thunderbird for your email that has RSS support too.
For Usenet groups the same. SABnzbd is by far superior to any other program. Combined with some other programs you can set up a command train that downloads, repairs, and extracts anything you queue up. Your weekly serial fully automated available as soon as it's out.
Using a full Windows install for just a news reader is ridiculous unless it pays you in gold for every message you read.

Yer, my hypervisor of choice is VMWare v16, will upgrade to 17 next full rebuild. If I could workout a viable type 1 where I had full hardware pass-through to allow swapping between M$ and Free systems so I could transfer seamlessly that would be awesome. But between slightly odd hardware, triple monitor, 4 sound cards /w 3 headphones + 1 bookshelf 2.1 not including monitor speakers, multiple keyboards/keypads. There are a few other peculiarities but those are the main ones.

As far as Linux goes my main flavours are Mint, Fedora/Redhat, SUSE and Kodachi. I've used others but those are the main ones.

I first attempted to use Linux in about '98 and have also used bare metal Unix System V, though that is going back. Over the last four+ decades I've had mixed successes with the various systems and environments available to me. I'd love to get more into retro emus (not the birds).

As far as platform specific software I do have a ton of native Windows stuff, productivity and gaming. If I can find a sensibly priced sparky to do the work needed I'll move many of these build across the house and into a 22RU enclosed cabinet. Physical and virtual alike. Out of sight and sound, the hypervisor yet tbd and I'll just remote in when need be, so local hardware/config won't matter.

I've likely only had one significant fault with each OS installed, two at the most. not a bad track record. I have seen some very interesting issues crop up here and there on personal and systems I've looked at in a professional setting, hardware, software and applications covered. I've diagnosed just about everything that can go wrong and has gone wrong at least once. One other option I could get into would be a dedicated machine on another desk. I have a 24" iMac (needs a new HDD) and more towers than I frankly have right to hold onto laying around.

I'm not shy of hardware, perhaps a spot to put it up however... If I don't have sight, sound and touch on my side I find my success rate of repairing a system goes down precipitously. I have to ask, what is the most random fault you have encountered?

what is the most random fault you have encountered?

An HDD with faulty RAM cache, when writing or reading from the drive, there was a 1 on 10E10 chance that bit 7 was flipped to 0. No Hardware test was able to detect the fault, but a simple file copy and compare would show the problem.

To this day there are still no hardware correction on hard drive cache or SSD cache. Would that drive have been in a RAID array the data would still have been corrupted.

Since then, I do all my work in VM on top of ReFS or ZFS whenever possible.

Unfortunately, Microsoft dedication to ReFS is underwhelming as more than one windows update lead to corrupted ReFS drives.

I reserve a place in hell for all storage manufacturer as that major issue with data integrity is still not addressed. There are integrity check and correction code at all steps (RAM, PCI, magnetic or flash storage but still nothing for the drive cache.

Can always relate to that sort of thing. lost a RAID volume to a bios issue that de-joined the array of drives. It was the early days of GUID partitions (4TB volume)so the recovery tools hadn't yet caught up. I did eventually get all the data back though.

When I can finally put together the hardware most of my bulk data and VMs will be rolled into a TrueNAS box running the equivalent of Raid60 (3x6+2) with ZFS and limited duplication. I will also run a rotation of tapes and cold hdds for stuff I really don't want to lose.

As for data integrity, I've never known why they can't have an equivalent to ECC in that chain. One of the most interesting causes for a fault in that chain was a partially failed cap (de)stabilising the CPU causing random bit flips every now and then. Eventually the system just refused to complete POST and just spat out a never ending meaningless series of beeps.

The real problem of ZFS is that you are stuck with BSD or Solaris

TueNAS is FreeBSD based and BSD is really bad at compression/encryption that translate to poor network and disk performance.

You'll have to use bhyve on top of zfs volumes. You'll need to limit zfs ram usage to have anything left for your VMs. But don't forget minimum ZFS ram is 8GB + 1GB RAM/TB of physical storage.

As far as I know, bhyve isn't mature, it still doesn't support overprovisioning of RAM and CPU overprovisioning was buggy and laggy as hell.

So better be sure of your workloads before investing in a TrueNAS hardware, you can't always add ram later.

the file server and hypervisor are physically separate machines though yes the TrueNAS box will have a small number of containers / jails whatever you want to call them. I'm familiar with the ZFS reqs and am planning ~256GB if not more. I've also looked into though not recently Linux based TrueNAS. Chatting on the TN forums if you're willing to sacrifice performance you can cut back on that GB:TB ratio. It's a long term project not likely to get off the ground before next xmas at the earliest.

You certainly are not stuck with BSD or Solaris for ZFS. OpenZFS runs on FreeBSD, Linux, and MacOS. There are some minor feature flags that may not supported on all ports, but things are converging quickly. MacOS may lag some, due to fewer developers.

I'm running ZFS with native encryption on Debian. I'm cautious, so I'm not booting from ZFS, but it is possible. I updated the kernel today, and apt automatically compiled the ZFS modules.

I know. Initially it was on FUSE with poor performance, now it's out of tree.

That mean any update to the kernel can break ZFS. And the kernel team won't care one bit.

Linus reply to such a break was:

If somebody adds a kernel module like ZFS, they are on their own. I can't maintain it, and I can not be bound by other peoples kernel changes.

So it's too experimental for me.

Linux has good bit rot protection with the btrfs file system but I was left with unmountable btrfs partition too many time to trust it fully.

Sure, if you roll your own kernels and ZFS versions, you'll break things. Linus is right, he won't maintain it. The OpenZFS people will, and tell you which which ZFS versions work with which kernel versions. If you use a distribution that supports ZFS and don't go off on your own, it just works.

I stick to EXT4 for all my storage. Might not be the fastest system and it does miss out on some handy functionality but oh so very reliable. I never had any data loss with EXT4, not even after several consecutive system crashes.

what is the most random fault you have encountered?

A very sporadic memory error on a Windows XP system, the only symptom was that XP occasionally told me the serial number was bad and *that* had serious side-effects.
In the end I ran memtest from a Linux CD and one of the tests showed the problem quite clearly. This must have been around 2004/5.

I think my most memorable Windows fault was someone who put an off-brand post-it note on an original 98SE disk that damaged the data layer and they couldn't re-install as a result.

But there was nothing 'wrong' with Yugo's till they banned tetra-ethyl-lead. Poor old T.M. Jr