Greg's recent activity
-
Comment on Python Foundation goes ride or DEI, rejects US government grant with strings attached in ~society
-
Comment on Making liquid nitrogen from scratch (an absurd amount) in ~science
Greg Link ParentI got the vibe that he ended up with a super high end system for about the price of a low end one, rather than the original plan of an equivalent low end one for cheap, and he was probably playing...I got the vibe that he ended up with a super high end system for about the price of a low end one, rather than the original plan of an equivalent low end one for cheap, and he was probably playing it up the a bit for the narrative (it’s been 15 years since I was last in an actual lab, but I remember liquid helium stuff being on a whole different level to liquid nitrogen so I’m guessing that hasn’t drastically changed).
But yeah, there were definitely a few places the sketchiness made me wince a bit, and I’m just hoping that was better managed off camera too.
-
Comment on What is happening to Japan? in ~tech
Greg Link ParentI think that dynamic actually made it one of the most interesting videos I’ve seen in a while. You can really hear the difference in his voice a couple of times towards the end as he slips out of...I think that dynamic actually made it one of the most interesting videos I’ve seen in a while. You can really hear the difference in his voice a couple of times towards the end as he slips out of “professional with a script” tone and into “I am genuinely upset” - I haven’t often heard creators just straight up say they hate their thumbnails and titles rather than putting a bit of even handed corporate spin around it, for example, and certainly not heard them say it with enough emotion that I fully believe they mean it.
-
Comment on In the early 1990s, Sweden faced one of the worst economic crises in its modern history – the lessons for other countries, especially France, deep in its own budget crisis, are simple, if not easy in ~finance
Greg Link ParentI’m not an expert on how the euro is administered, so I could be missing something here, but isn’t it just transferring that authority from national to supranational level, rather than handing it...I’m not an expert on how the euro is administered, so I could be missing something here, but isn’t it just transferring that authority from national to supranational level, rather than handing it to the private sector?
-
Comment on Amazon ordered to pay $20K after British Columbia customer says package never arrived in ~tech
Greg Link ParentYeah, you’re probably right about the legal outcome, and I’m always disappointed (as it sounds like you are too) when “throw it in the 10,000 word T&Cs that nobody will ever see” is given as an...Yeah, you’re probably right about the legal outcome, and I’m always disappointed (as it sounds like you are too) when “throw it in the 10,000 word T&Cs that nobody will ever see” is given as an acceptable solution.
I was mainly thinking that a photo would help prevent it going as far as the court in the first place - it shows whether it’s a simple mistake and helps the customer fix it, or whether they need to make a police report for theft. I guess I’m also thinking that the absence of a photo points the finger more at Amazon, even if the presence of one wouldn’t definitively absolve them; less from a legal perspective and more just from a “this seems sketchy compared to my personal experience” perspective.
And yeah, they definitely have OTP infrastructure in place here (UK), but it’s only used very rarely. I’ve seen it maybe two or three times ever, and only on high value but physically small electronics (CPUs, large SSDs, but not on a GPU, or a phone, both of which I would have expected to fit that pattern). I’m sure there’s some complex heuristic behind what does/doesn’t get flagged, probably a pretty good one given how much data they have, and I get that it adds a lot of friction so they don’t want to enable it too broadly, but it seems wild that when they also have a “don’t refund this specific order for this specific customer” heuristic, that doesn’t trigger a OTP!
-
Comment on Anthropic aims to nearly triple annualized revenue in 2026, sources say in ~tech
Greg Link ParentIn theory it’s not that wild for a rapidly growing company to be judged on recent performance - especially if they’re working with longer B2B contracts, in which case the whole bunch of deals...In theory it’s not that wild for a rapidly growing company to be judged on recent performance - especially if they’re working with longer B2B contracts, in which case the whole bunch of deals you’ve just signed last month are already agreed to continue for the next 12+ months anyway, and it’d be more misleading to average over the 11 prior months where those revenue streams didn’t exist than to project them forward into the year you already know they will exist.
But in practice, anything remotely reasonable in business has been stretched far past its breaking point, the pieces have been sold off for scrap, and then 10x their actual value has been recorded as a tax loss. The concept can be sound, but I wouldn’t trust the people using it as far as I could throw them.
-
Comment on Amazon ordered to pay $20K after British Columbia customer says package never arrived in ~tech
Greg Link ParentThat's a very interesting nuance about the photographic evidence, but I'd say Amazon are weasel wording it a bit - there's a middle ground on the spectrum between "no photo" and "fully...That's a very interesting nuance about the photographic evidence, but I'd say Amazon are weasel wording it a bit - there's a middle ground on the spectrum between "no photo" and "fully identifiable photo of an individual". I'm not in Canada, so I can't speak for PIPA, but I am in a location subject to GDPR and it's totally standard for delivery drivers to take a photo of the package in the customer's hands or in the doorway if they deliver it directly to a person. They do always take a picture of the package specifically, not the person, which suggests they've probably been trained to do it that way.
Even a photo on the porch does quite a lot - it shows that they got the right house rather than a neighbour who's still within the GPS margin of error, it shows that the driver actually got out of their van rather than just stopping to satisfy the GPS and then driving off because it's cold or rainy or they don't want to bother lifting heavy boxes (this is an actual fight I've been having with a non-Amazon delivery company recently, where I thankfully have video evidence on my side), it shows that they did leave it on the porch rather than "helpfully" hiding it in the recycling bin an hour before it gets emptied or anything like that, it shows that the label they scanned was on a box of the correct approximate size, etc. etc.
Like you rightly say, OTP is the only way to really ensure that a specific person received the package (or at least received a package - even then there are edge cases like a driver handing over three boxes, asking for the code, and the customer not realising the code applied to a fourth box), but a photo rules out almost all honest mistakes. It narrows it down to porch pirates, customer fraud, or driver theft after the photo was taken.
-
Comment on Amazon ordered to pay $20K after British Columbia customer says package never arrived in ~tech
Greg LinkI used to decry the fact that companies would delegate decision making to automated systems with no means to escalate to a human for review - I'm actively in favour of automation in general, but...I used to decry the fact that companies would delegate decision making to automated systems with no means to escalate to a human for review - I'm actively in favour of automation in general, but it seemed insane to me that they'd accept the machine's decision 100% of the time, rather than taking it as a 99% efficiency gain and letting a human correct the 1% of obvious failures when they're flagged.
What I'm seeing that seems much worse recently is the humans sticking just as rigidly to these absurd decisions as the machine would have done. These are all wildly varying in severity, but they seem like the same root problem of rigidity over common sense: just this week there's been guns pulled on a teenager because a bag of doritos was flagged as a gun, a £150 fine for tipping half a cup of coffee down the drain (yes, they relented, but only after two days of internet mockery - they originally doubled down even when contacted by the press), and now an Amazon case that cost $20,000 of lawyer time rather than being a ten second conversation along the lines of "Are we sure that was delivered? Well, there's no photo, so I can't be certain... Right, refund it, what's next".
-
Comment on Amazon ordered to pay $20K after British Columbia customer says package never arrived in ~tech
Greg Link ParentIt could be “suspicious” against the customer, it could be a statistical fluke, or it could be a sign of a repeated issue on Amazon’s part: a software issue incorrectly labelling the address, a...It could be “suspicious” against the customer, it could be a statistical fluke, or it could be a sign of a repeated issue on Amazon’s part: a software issue incorrectly labelling the address, a confusing building layout that new drivers aren’t familiar with, a habit of leaving things in unsafe locations, even a delivery driver taking them.
Surely the onus is on Amazon, who has every means to create proof that they did deliver, to show that the customer is at fault for this “suspicious” pattern? Because it’s just as likely for that suspicion to point right back at Amazon, either through malice or incompetence, and the customer has no means to definitively prove that something wasn’t delivered.
The lack of a delivery photo seems like a big enough oversight to side with the customer by default, and if Amazon already know an account is flagged for “return abuse” they have plenty of options to create additional proof rather than summarily denying refunds. There’s already a mechanism in place to require a one time code for higher value items, they could trivially enable that on any item that would otherwise not be refunded. Hell, they could even make it a click through: “Use a one time code or accept the risk that this is non-refundable even if it doesn’t turn up, your choice” - it’s not great UX, but neither is taking a customer’s money for something that didn’t arrive, so I can only assume that most people with an honest delivery problem would be actively happy to solve it this way.
-
Comment on The majority AI view in ~comp
Greg LinkIt’s a breath of fresh air to see an article like this. It seems like everything I read is billionaire hype or (understandable, justified, but often technically misguided) backlash against the...[W]hat they all share is an extraordinary degree of consistency in their feelings about AI, which can be pretty succinctly summed up:
Technologies like LLMs have utility, but the absurd way they've been over-hyped, the fact they're being forced on everyone, and the insistence on ignoring the many valid critiques about them make it very difficult to focus on legitimate uses where they might add value.
If we were to simply listen to the smart voices of those who aren't lost in the hype cycle, we might see that it is not inevitable that AI systems use content without the consent of creators, and it is not impossible to build AI systems that respect commitments to environmental sustainability. We can build AI that isn't centralized under the control of a handful of giant companies. Or any other definition of "good AI" that people might aspire to. But instead, we end up with the worst, most anti-social approaches because the platforms that have introduced "AI" to the public imagination are run by authoritarian extremists with deeply destructive agendas.
It’s a breath of fresh air to see an article like this. It seems like everything I read is billionaire hype or (understandable, justified, but often technically misguided) backlash against the billionaire hype - and as someone who works in scientific ML, a field that’s firmly swept into the “AI” catch all but has nothing to do with LLMs, it’s nice to hear a voice of moderation.
And the author’s absolutely right, if we treat it more like the “normal technology” that it is, we might just break this idea that it’s synonymous with Sam Altman and Mark Zuckerberg’s bullshit, defuse some of the backlash, and have an opportunity to make use of it in a positive way.
-
Comment on Not sure if coincidence or I should give up (on USB flash drives) in ~tech
Greg Link ParentI tend to compare on IOPS if I'm looking to get an idea of performance between drives - it gives you a single number to compare when scrolling through a list, sequential performance is rarely a...I tend to compare on IOPS if I'm looking to get an idea of performance between drives - it gives you a single number to compare when scrolling through a list, sequential performance is rarely a bottleneck, and I find IOPS is a better real world measure of what to expect than trying to balance the relative impact of controller type, DRAM cache, SLC cache, and NAND type. Something like the WD SN7100 is a good example that looks like it should be middling performance based on the lack of DRAM, but actually comes close to the Samsung 990 Pro in most real world tests even though it's 30% cheaper, which is a lot easier to see if you're comparing price/IOPS.
I'll go deeper and look up actual benchmarks of a specific drive if performance is critical, but like @bitwaba said, even the basic performance numbers are probably not worth worrying about for a pseudo USB stick. There's a whole other can of worms around USB-NVMe bridge chips there if you're actually trying to maximise performance, too, so unless you're planning on booting from it or running a database on there or something like that, almost anything on the market should be fine here.
-
Comment on Not sure if coincidence or I should give up (on USB flash drives) in ~tech
Greg Link ParentI buy a fair amount more drives than the average person, and I’d have no major reliability concerns buying from any of the brands you’ve likely heard of (Crucial/Kingston/Samsung/Western...I buy a fair amount more drives than the average person, and I’d have no major reliability concerns buying from any of the brands you’ve likely heard of (Crucial/Kingston/Samsung/Western Digital/Seagate/Sabrent/Lexar/probably a couple of others I’ve missed). As long as you don’t get a counterfeit (looking at you, Amazon), and don’t buy something from an all-caps-no-vowels brand that’s sourcing their flash from the waste pile behind a chip fab (looking at you again, Amazon third party sellers), you should be fine.
You might not be getting great value right at the bottom end of the market - random write performance in particular (copying thousands of small files, rather than a few big files) can be 500% better for a 50% price increase - but that’s comparing to other SSDs. If you’re replacing a USB flash drive even the cheapest SSD should be a decent improvement, and if you’re only copying a few GB at a time you’ll never have time to notice it’s “slow” regardless of write patterns.
I’m normally a bit more concerned about performance than is relevant here, but for what it’s worth most of what I do end up getting is Samsung or Crucial (which is just Micron with a consumer-friendly hat on) and they’ve always been solid - that’s also less a question of brand loyalty and more because the two largest manufacturers of actual flash chips seem to do a good job of hitting the price/performance sweet spot for assembled drives! I grabbed a couple of WD drives the other day because they were on sale and didn’t think twice about it, for example.
-
Comment on What's a quantum computer? in ~tech
Greg (edited )Link ParentThey do exist (one of my closest friends programs them), but they’re somewhere around where classical computers were in the 1950s-60s. Big pieces of research equipment with teams of academics...They do exist (one of my closest friends programs them), but they’re somewhere around where classical computers were in the 1950s-60s. Big pieces of research equipment with teams of academics simultaneously working out how to build them and how to use the things they’ve just built.
I’ll never say never on them somehow finding their way into personal computing - science and tech seems to have a strong track record of finding uses that nobody expected for things as they advance - but realistically I’d be surprised to see it happen in our lifetime, and if anyone’s trying to pitch end user applications for these things in the next couple of decades they’re probably lying.
That said, it’s not quite at the LHC level either: that’s a global one-off with pure science as its goal, so I’d put quantum computers somewhere in the middle. Maybe something like an injection moulding machine, or a combine harvester - big, expensive, specialist equipment that does a particular thing for a particular job. Even if you or I could somehow spend six or seven figures on one, there’d be no reason to: I don’t need to mould 150,000 identical widgets, and my cheap 3D printer is better for making the one off bits I do need; I don’t have a field to harvest, and farm equipment is a terrible option for driving to the shops.
The kind of companies that would buy a room-sized IBM machine in 1962 may well have a reason to buy a quantum computer in 2032, but they’re unlikely to be something the rest of us really encounter any time soon. Also how in the hell does “plausible date a few years from now” get me to twenty thirty goddamn two?!
-
Comment on Looking for feedback on a homelab design in ~tech
Greg Link ParentVery glad it was helpful! And I've definitely come across that same challenging feeling when searching for info, so I'm pleased to have had a reason to get it all written down in one place....Very glad it was helpful! And I've definitely come across that same challenging feeling when searching for info, so I'm pleased to have had a reason to get it all written down in one place.
Totally fair on the JBOD side; I'm limited by rack space, and even more limited by floor space for another rack, so that's probably colouring my thoughts even when I'm trying to be more general. I think I had a bit of an impression that you were worried about overflowing a 45 bay shelf, but it doesn't sound like that's the case!
Similar on the VDEV layout, I think I'm naturally assuming large drives just because I'm always keeping compactness in mind, so when you say 11 drive VDEVs starting with two and scaling to four or eight, I hear half a petabyte to start, possibly growing to 2PB. That's a pretty serious installation even by large organisation standards (cough Korean government), and a solid few thousand euros/dollars/pounds if you need to add a VDEV, but if you're looking at old 4TB or 8TB SAS drives it's a whole different ball game. I'd probably still lean towards smaller VDEVs with bigger drives just because they're likely to be newer, but that's likely just bias creeping in on my side.
That failure visualisation site is cool, by the way, I hadn't come across it before! I think what I'd say there is just to keep in mind the externalities. You're not worried about the probability of failure per se, you're worried about the probability of multiple failure in the day or two it takes for a replacement drive to be delivered, and even with a "bad" VDEV layout the numbers are so low that you want to be looking at them alongside the chances of losing the server as a whole to a faulty PSU or burst water main or lightning strike on the power line. A single hot spare takes delivery time out of that equation and stacks the odds even more your way.
There's certainly no harm in thinking about drive failure rates, or in optimising against them to a degree, but (and I say this as someone very prone to bikeshedding, who needs to hear it myself!) if you get a multiple simultaneous drive failure it's more likely to be because a meteorite hit your server and you need to restore from offsite backups anyway.
-
Comment on Looking for feedback on a homelab design in ~tech
Greg Link ParentIt's a good question! Everything actually does get written twice for sync writes - first to the ZIL (ZFS intent log, this is the log you're separating out with a SLOG drive, i.e. a separate log),...It's a good question! Everything actually does get written twice for sync writes - first to the ZIL (ZFS intent log, this is the log you're separating out with a SLOG drive, i.e. a separate log), and then to the actual ZFS filesystem properly. Whether or not a SLOG drive is being used, the ZIL exists so it can be written as fast as the underlying storage device will allow, so that the sync write can be acknowledged and the program that's waiting on it can continue - it's basically a dumping ground for ZFS to put raw data into and confirm the write without having to stop and think about it, before then figuring out where it goes in the pool and writing it properly afterwards.
If the ZIL is on a spinning disk, the head has to flick back and forth constantly between the ZIL area of the disk and the actual zpool area of the disk, and because sync writes are latency sensitive it really is constant - it keeps getting pulled back to the ZIL millisecond-by-millisecond as new writes come in, while still trying to keep up with the overall pool writes (including the "real" writes for the data that was temporarily dumped in the ZIL) which pull the head back to a different area of the drive.
A SLOG drive removes this issue of a physical head being pulled back and forth on the same device, and on top of that means that the only writes going to the spinning disks at all are clean, ready-organised sets of blocks that get flushed every few seconds - as opposed to many thousands of latency sensitive writes that need to be acknowledged immediately even if they don't nicely line up with the underlying filesystem state. So you've got 1x the data written to spinning disks rather than 2x, which is a big win to start with, and then you've got something like 0.1x or 0.01x the number of write requests to get that same amount of data into place because they're being batched efficiently.
And I can say from personal experience that it's really noticeable! I had an absolute "oh shit" moment when I installed a new batch of HAMR drives into a machine and it became so loud I could hear it in another room - I was worried they just ran that much louder than the non-HAMR ones I was replacing, and I was going to have to figure out some kind of janky soundproofing, but it turned out I'd just messed up the sync config. Without the constant stream of ZIL writes hitting them, the sound is barely noticeable even close to the machine, and if you do listen out for it it's a little "blip" every few seconds rather than a continuous (much louder) rattle.
Good info on SLOG and ZIL here, if you're interested: https://www.truenas.com/docs/references/zilandslog/
-
Comment on Looking for feedback on a homelab design in ~tech
Greg Link ParentAre you thinking of non-ZFS filesystems when you're talking about RAID? I can't imagine it'd be particularly convenient to run a NAS with the shares limited to the bounds of single hard drives,...Are you thinking of non-ZFS filesystems when you're talking about RAID? I can't imagine it'd be particularly convenient to run a NAS with the shares limited to the bounds of single hard drives, but as soon as you're spreading data across multiple drives you're in a situation that one drive failure loses everything if you don't have RAID(Z) or mirroring set up. Proper offsite backups are a necessity, no question, but I'd consider some level of drive redundancy a necessity as well otherwise you'll need to do a full restore (and deal with however many hours or days of data loss since the last backup) every time a drive fails.
I agree with you on infrastructure as code, I've never regretted the up-front time cost of having my configuration in one place that I can review/edit/diff it, although I'd want something state-aware like terraform or pulumi rather than ansible. After putting in that up-front time cost, I like to be able to have the software guarantee that the system state matches what the config file says, and that the updated state after making changes matches in turn, rather than just using it as automation for a series of steps.
-
Comment on Looking for feedback on a homelab design in ~tech
Greg LinkThis turned out way longer than I intended, but I hope it's useful! There are a ton more details and nuances floating around in my head, so absolutely feel free to ask if there are specifics you...- Exemplary
This turned out way longer than I intended, but I hope it's useful! There are a ton more details and nuances floating around in my head, so absolutely feel free to ask if there are specifics you want to discuss more.
I've had to do a kind of absurd deep dive into ZFS and the hardware/OS config/general NAS setup to run it recently, so happy to get as far into the details as you'd like! Something I've really found when searching is that a lot of the accepted wisdom is quite dated, even when it's being repeated in newer discussions - I've consciously gone against that in a few places, so if something doesn't seem to add up just ask, 50/50 whether it's deliberate or whether I've just made a mistake.
JBOD with either 90 bay or 45 bay storage
Roughly how much data are you expecting to store? This seems like a lot, especially when 100TB usable is five or six 28TB drives (the currently optimal price-per-TB). You might want to go 16TB or so just to give yourself a bit more granularity for VDEV configuration, but even then you're pushing 200TB usable space in a 4U case alongside the NAS server itself, no JBOD needed.
For what it's worth, you want to be buying the recertified with warranty Exos drives - redundancy is one of the major purposes of a ZFS setup, and no individual user is buying enough drives to be a statistically significant sample anyway, so as long as you're covered against DOA and early failure it's really not worth worrying about the difference between new and refurb/recertified. I've even seen some people suggesting that recertified have a lower failure rate due to the extra rounds of testing, although I haven't seen enough data to corroborate that.
Im tentatively planning for an 11 wide Raidz2 vdev configuration. This would hopefully scale to 8 vdevs with 2 hot spares or 4 vdevs with 1 hot spare.
If you're thinking of more drives for better throughput, I'm going to make a potentially controversial point and say that (within reason) you shouldn't be worrying about the HDD speed on a modern NAS. SSD and RAM caching are multiple orders of magnitude faster, and cheap enough that it's just not worth trying to optimise the bits of the system that are always going to be limited by a physically moving arm and platter.
AMD ryzen CPU and motherboard
64 or 128GB ramZFS loves RAM, so more tends to be better, but you can always start at 64GB and see if you actually hit any bottlenecks from cache misses first.
You'll probably want to run the memory at lower than its rated speed, too. This isn't a problem - you will absolutely never notice a difference in memory speed when serving files unless you're also running $10k in networking hardware - but Ryzen only has two memory channels, so four DIMMs is running two per channel (unless you're planning on using 64GB DIMMs, but last I checked they were way too expensive compared to 32GB) and they tend to get unstable if you don't knock the speeds down a bit.
You'll also want to consider the PCIe layout of your motherboard, and especially look out for one with bifurcation support, because that'll let you drop in a whole lot more NVMe drives as and when you need them.
AMD epyc and motherboard if I can get my hands on a less expensive one
There are really good bundle deals on eBay at the moment for a Supermicro H11SSL-i or H12SSL-i along with RAM and an Epyc Rome (7002 series) or Milan (7003 series) CPU. I bought one a month or so ago and it's been great so far - I'd say it's worth paying a little extra to go for Milan or newer, Rome is getting pretty dated at this point. You can also save a little by getting a Gigabyte board rather than Supermicro, but everything I could find said they're incredibly picky about PCIe devices, whereas the Supermicro ones just work (which I can attest to). Shipped from China, which is the only thing that might be an issue if you're in the US.
I know Proxmox can do zfs right out of the box so I know I don’t need the TrueNAS server, but splitting it this way just seems more flexible. Is this a realistic setup or would it just be better to let Proxmox do everything?
Separate NAS is always a good idea in my opinion. Proper separation of concerns and the ability to add/repurpose more machines later is more than worthwhile for a setup like this.
You could always run Proxmox on the NAS box as well, just with a single TrueNAS VM inside it, if you want the extra layer of abstraction for system backups etc. - I'd consider this if using TrueNAS, but I'm using NixOS so I'm less worried about losing system installs or OS config. If I were going down that route I'd still let TrueNAS handle the ZFS side, rather than giving it to Proxmox. But honestly as long as you have config backups for TrueNAS I'm probably overengineering things by even suggesting virtualising it.
Does anyone have experience creating NFS shares in TrueNAS for mounting in Proxmox? I would be interested in thoughts on performance, and stability among any other insights.
If you're sharing sensibly sized files (few megabytes to tens of gigabytes) I'd expect this to be fine - large numbers of much smaller files are where you start getting problems. I've found that NFS pretty much just works, and has decent throughput, but does add enough latency that you'll notice it when you need to do something like recursively list a directory tree with a few tens of thousands of files and subdirectories.
Is SLOG and/or L2ARC necessary for this setup? What capacity and configuration would be best?
L2ARC: yes. You don't want to be waiting for hard drives if you can reasonably avoid it, and you definitely don't want to be waiting for hard drives to look up metadata. Ideally you'll be hitting ARC rather than L2ARC more often, but even 64-128GB in comparison to HDD sizes means that things definitely will be pushed out relatively frequently, and 2TB NVMe really doesn't add much cost in the context of the hardware you're talking about here. If you find cache misses are a bottleneck once you get a feel for things, it's easy to add more NVMe drives later as long as you have the PCIe lanes for it.
You'll see a lot of people online saying SLOG is unnecessary because it's only required for sync writes (async writes are acknowledged as soon as they're cached in RAM, and are flushed to persistent storage every 5 seconds or less by default, whereas sync writes are only acknowledged after they're in persistent storage). What they don't tend to mention is that NFS defaults to using sync writes for everything!
Your options here are:
- Pass the
asyncflag when mounting the NFS shares. In the worst case this can cause data loss in case of network disconnection, power outage, or system failure, because NFS on the client system will confirm writes before they've been written by the server. - Mount the NFS shares synchronously, but force ZFS to write them asynchrounously. This removes the network as a failure point, because NFS waits for the NAS to confirm writes, but allows ZFS to send that confirmation while they're still cached in RAM on the server. Worst case risk is data loss of up to the last 5 seconds of writes in the case of a power outage or system failure on the server, but not the network. Bear in mind that "risky" here is by ZFS standards, which are still more reliable than most filesystems' defaults.
- Keep everything running synchronously without a SLOG drive. I wouldn't do this - sync writes on spinning disks take a comparative eternity, and it also leads to the drives constantly thrashing and creating noise and vibration - but you can always test it out if you want. If you have a known workload with large files and correctly tuned block sizes it might just about be acceptable. Maybe. This is safe against data loss in all circumstances, with writes only being acknowledged after they're in persistent storage.
- Keep everything running synchronously, use a SLOG drive. Same safety guarantees as the previous point, but much faster because you're caching those writes to an SSD - persistent, unlike RAM, and responsive, unlike HDDs.
But there is a major caveat on the last point: consumer SSDs don't actually write things persistently in the way most people expect. The faster ones have an internal DRAM cache, which is itself volatile before the drive controller can flush it to the NAND. For most consumer-grade drives, this means the latency on SLOG writes is a lot higher than you'd expect it to be, because ZFS is waiting for the NAND write rather than the DRAM write: still a decent amount faster than a HDD, but nothing like the snappiness you get from a "normal" filesystem that just assumes the drive's DRAM cache is safe enough to consider successfully written. For some drives, the controller will falsely report to ZFS that data was written when it's actually in the cache: this is much worse, because ZFS can deal with known data loss or corruption, but isn't built with an expectation of drives giving incorrect information. A power loss on the latter type of drive could theoretically leave ZFS itself in an inconsistent state.
If you're going for a SLOG drive, make sure it's an enterprise grade one with PLP (power loss protection) - those can correctly acknowledge writes even when they're in the DRAM cache, because they have an onboard capacitor that will allow the controller to flush the DRAM to NAND even in the case of total power loss, and without any interaction from the filesystem or OS. You'll see a lot of discussion about Optane drives, which were basically this setup with an unusually large (for the time) cache, but those haven't been in mainstream production for years and are now even hitting the end of extended enterprise support. Modern drives have large enough caches and fast enough flash that I don't think it's worth considering Optane, although if another manufacturer picked them up and made a modernised and supported version I'd be interested!
Micron and Kingston both make M.2 2280 NVMe drives with PLP, or you can look at used 2.5" U.2 enterprise drives and get an adapter to connect one to a PCIe or M.2 slot. Either way, you'll only need a few tens of GB for a SLOG drive, so pretty much anything on the market will be fine in terms of size - since SLOG is inherently a write-heavy workload, you can consider the extra space to be overprovisioning for wear levelling.
-
Comment on Amazon Web Services outage impacts in ~tech
Greg Link ParentOhhhh, that's why Signal says I'm offline! Glad I don't have to go digging into firewall settings or anything.Ohhhh, that's why Signal says I'm offline! Glad I don't have to go digging into firewall settings or anything.
-
Comment on How can I combine several ranked lists into one mega list? in ~comp
Greg Link ParentI agree - for all the times I’ve seen spreadsheets used for things they shouldn’t be, this one seems like they’re the right tool for the job. I’d be thinking average of the item’s position across...I agree - for all the times I’ve seen spreadsheets used for things they shouldn’t be, this one seems like they’re the right tool for the job.
I’d be thinking average of the item’s position across lists as the score to sort by, either divided by the number of lists a given item appears in, or ranking an item
len + 1for any list it doesn’t appear in, depending whether you want to treat non-appearance as neutral (i.e. these are categorised collections and you don’t expect Pokémon to come up on the top ten racing games in the first place) or negative (if it didn’t come up on the top ten best games of all time, it’s presumably lower ranked than all ten). -
Comment on A quarter of the US Centers for Disease Control and Prevention is gone in ~society
Greg (edited )Link ParentRight now the constitution is being ignored in every meaningful way and nobody’s enforcing any of the existing checks, so I think it’s more a question of reclaiming democracy first, and then...Right now the constitution is being ignored in every meaningful way and nobody’s enforcing any of the existing checks, so I think it’s more a question of reclaiming democracy first, and then cementing it.
If the Heritage Foundation still exists and wields power, there’s no point in the convention not just because they might co-opt it, but because even an absolutely perfect amendment that anticipates and redirects every possible problem would still be nothing more than words on paper. Words without enforcement won’t get the US back on track.
Amending the constitution isn’t going to be a fix, it’s going to be the thing you do after fixing the problem, in an effort to stop it happening again.
[Edit] For complete clarity, I’m agreeing with @vord here: the states coming together in a way that’s been unprecedented in living memory to say “we are a stable, democratic country that rejects fascism” is one of the few ways America will be able to draw a line under what’s happening right now, and signal that change in a way that the world can trust. But it will be a signal and a mechanism to protect against the same thing happening a second time, it won’t be the means by which the change happens.
Yup. I’m glad to see the python foundation making it clear that they are taking a stance as a matter of principle, but this hits the nail on the head even if it were a totally amoral business making the decision.
You can’t make significant financial decisions based on a buzzword from a government that’s shown zero consistency and no willingness to act in accordance with the law. You really, really can’t make them if that government also has the power to take the money back after you’ve spent it if you do something to piss them off. Or if the commander in chief happens to see an AI generated video that pisses him off and has a word in it that sounds vaguely like the name of your organisation, for that matter.
This is why corrupt states don’t work. It’s not just a moral issue (but it is a big moral issue too), it’s a practical one. Business only works if there’s trust. Finance only works if there’s trust. And if the people enforcing that trust are the ones trying to fuck you over, the whole system crumbles.