Lost a Node today

I lost a node the other day due to a disk crash

Node ID: 118wwYmg4sUJsrxdBWmbdxN6P9UzJRUjffrHGzFi6WUNHJP2EX

Its not coming back - disk loss was sudden (no warning) and total

Did you take it out back and shoot it?

5 Likes

Surely it’s a coincidence but, after 5 years without any disk issue, the last 2 months i had segment faults in 2 disks. One of them was 1 year old :smiley: Luckily i could copy 99% data but it’s a pain

Time to pour one out for the fallen soldier.

I thought I had lost my original node’s disk a while back, fortunately it was just a cabling problem on the disk.

And this is precisely why SMART tests are a massive waste of energy/performance/time. Nobody can predict the future.

Disks either read and write data or not. Once they don’t — they get replaced.

I myself never run any kind of smart tests. When filesystem reports IO errors — disk gets replaced. Easy. If there are no errors reported and disk does what it supposed to — I don’t care what smart says. I’m not replacing disks prematurely.

How do you know the OP observed SMART values?

Gone to the round filing cabinet

1 Like

I do - but as I said - no warning. Disk just stopped responding. Placed in another machine - and no disk present - its dead Jim. About 6 years old

Node was full, 9TiB full (or thereabouts) on a 12TB disk - and had just gone to 5TB of trash too.

I run the most smart tests when I get a new drive. Because I rarely get ‘new’ drives but instead used or refurb drives. I generally do one pass of writes (badblocks or a slow format) and then an extended smart test. If there are problems I return or retire it.

But if I start getting weird errors I may take the drive out of commission but run tests to figure out what’s up. That drive that I thought was a dead node was actually just janky cabling problems.

Interesting. When I was foolishly buying new devices I would too test them. I don’t bother with used/recertified drives — they are in the middle of the bathtub curve, they have outlived the early failures.

It’s so frustratingly difficult to find cables that don’t suck. Pretty much everything on Amazon is trash. Some brands, (like cablematters recently) sell marginally less crappy junk, but it’s still crap. The only avenue is old cables from decommissioned servers and anything SAS related — because they those were not made for consumer market.

And power supplies. Situation with power supplies is dire. Just look at what shite is “upscale” Synology shipping… but that’s an EE rant for another day.

I’ve been recently getting components for a new desktop and got this list recommended. My current desktop’s PSU is rated as “avoid” :sweat_smile: but the list does look solid.

1 Like

With PSU I streamlined my approach completely: I just buy 80 plus titanium, (or platinum, in rare cases, when titanium is not available). Manufacturer does not matter to a degree: I won’t buy Gramplo or barmatron or floorgoo from Amazon of course, but if, say, ableconn makes titanium psu — that’s ok with me.

Im not after power savings per se, even though those are a nice bonus; it’s because if they bothered to add circuitry and tune it to maintain high efficiency at those 3 or 4 load levels, they have very likely also did not skimp on transient suppression, noise, ripple, and a miriad of other things that may or may not being tested in the article you’ve liked, which I skimmed but have yet read.

There are so many ways to cut corners when designing a power supply it’s not even funny. It will pass all the tests for a while too.

1 Like

If you buy from brands that no one heard of, don’t be surprised when you have problems.
Just stick with the big names like Corsair, EVGA, SeaSonic, Thermaltake, etc. Don’t just buy what Amazon feeds you. The ratings on cheap brands are meaningless. The chinese lie about specs like they breath air.

“Big brands” rebadge other OEM units, especially entry level ones, all the time.

Also seasonic and corsair in the same bucket… is weird. Seasonic is known for their power supplies. Corsair is known to rebadge anything that moves on earth. “Their” products are mostly turds wrapped in marketing fluff.

I understand the youngsters among us easily forget (or don’t even know), but for the OGs
out there (all 5 of us), they’ll remember Jow Gerow (of jonnyguru.com fame, aka the best PSU reviewer of the Old Golden Era of Computing), who is now a director for PSU R&D at Corsair.

TL;DR: Corsair develops their own PSUs.

I never said they don’t. I said Corsair rebadges a lot products, including PSUs. Maybe they also develop excellent ones — the point is you can’t just go by brand name.

Of course rebadging is a common practice in all industries, but when a big brand slaps it’s logo on something, you have some assurance; first of you have their warranty, second they don’t just take a crap from the bunch and rebrands it; they make sure the crap meets some requirements, because if something happens, the fingers are pointing to them not to the original crap manufacturer. So… I radder buy rebranded than no branded.

2 Likes

Not really. The only assurance you have is that the brand deemed it more profitable to rebadge the device to fit their product line and extract more money from customers.

This is not black and white. It does not have to burst in flame. You can have a power supply mostly work, and if it fails to perform users and vendor will chuck this out on the external factors (but you had a transient, but your UPS did not switch fast enough, but your mains power browned out, but you have continually used 90% of capacity, but you ran it in hot or cold environment, etc. ) while better supply would have handled that without you noticing.

The golden sample maybe met some requirements when the contract was signed.

And you probably are overestimating the rigor of testing they do. And then, it’s still cheaper to make a power supply to pass tests than to perform well for the user.

As an example you can look at performance of 80 gold power supplies from weird vendors — they will be under threshold at the mandated power levels and all over the place in between. Better power supplies will provide adequate efficiency across the board. Both “pass the tests” but one is shit, and the other may be not.

(If you notice extra negativity towards Corsair — it’s because from past experience I would not touch anything they sell with a 10 foot rotten stick, but the same logic applies to other power supply vendors too, from USB-C bricks to UPS inverters. It’s very easy to cut corners and there really is no reason not to for a for-profit company with target audience that won’t know any better)