On 21/07/2024 21:28, druck wrote:

On 21/07/2024 11:47, Ahem A Rivet's Shot wrote:

On Sun, 21 Jul 2024 10:44:03 +0100
The Natural Philosopher <tnp@invalid.invalid> wrote:

Oh indeed. My new server will feature two SMART enabled SSDs...one a
mirror of the other.
I am not interested in RAID. RAID increases availability, but does not
archive data

    You have a mirror - that's RAID. RAID is about smoothly surviving
drive failures. With any storage system there are two important factors -
mean time to data loss and probability of data unavailability.

Ignoring whether its RAID or not, mirroring will protect you against a
random failure of one of the drives, which was more useful in the
spinning rust days when random mechanical failures were an issue.

I agree.

With SSD, write life is the main issue, and if you have two identical mirrored drives, you may find any write life issues, which are not
random, occur at exactly the same time.

But never at exactly the SAME time.

Remember the primary drive gets written to all day long as stuff like
this post is downmloaded, read and deleted.

The secondary gets a once a day rsync,

And, whilst I have never had an SSD wear out fromn writes in the past 8
years I have had one fail in a quite different way shortly after purchase.

So with any type of mirrored arrangement, make sure they are different
makes or models of drive, so it is less likely they fail together.

They are not subject to the same usage pattern, and they are not made
from the same components.

As long as they dont fail within 24 hours of each other


--
In a Time of Universal Deceit, Telling the Truth Is a Revolutionary Act.

- George Orwell
--- SoupGate-Win32 v1.05
* Origin: Agency HUB, Dunedin - New Zealand | Fido<>Usenet Gateway (3:770/3)

On Sun, 21 Jul 2024 09:29:51 +0100, The Natural Philosopher wrote:

On 21/07/2024 09:05, Lawrence D'Oliveiro wrote:

On Wed, 17 Jul 2024 10:35:01 +0100, The Natural Philosopher wrote:

On 17/07/2024 02:30, Lawrence D'Oliveiro wrote:

On Tue, 16 Jul 2024 10:31:11 +0100, The Natural Philosopher wrote:

On 16/07/2024 01:57, Lawrence D'Oliveiro wrote:

SMART isn’t much use, anyway. I test my storage devices for actual >>>>>> I/O errors.

That's what you use SMART *for*.

No, I test doing actual I/O.

So does SMART.

No, it extrapolates from its internal firmware behaviour. It tries to
predict failures before they happen.

No it doesn't. It predicts ...

s/predicts/tries to predict/. It’s not a prophet, you know.

And it only catches about 30% of failures.
--- SoupGate-Win32 v1.05
* Origin: Agency HUB, Dunedin - New Zealand | Fido<>Usenet Gateway (3:770/3)

On Sun, 21 Jul 2024 09:31:03 +0100, The Natural Philosopher wrote:

On 21/07/2024 09:06, Lawrence D'Oliveiro wrote:

This is why you have redundant systems. That’s how the pros do it.

No. Its why you use SMART.

Companies whose business it is to ensure data integrity do not rely on
SMART.
--- SoupGate-Win32 v1.05
* Origin: Agency HUB, Dunedin - New Zealand | Fido<>Usenet Gateway (3:770/3)

On 24/07/2024 01:33, Lawrence D'Oliveiro wrote:

On Sun, 21 Jul 2024 09:29:51 +0100, The Natural Philosopher wrote:

On 21/07/2024 09:05, Lawrence D'Oliveiro wrote:

On Wed, 17 Jul 2024 10:35:01 +0100, The Natural Philosopher wrote:

On 17/07/2024 02:30, Lawrence D'Oliveiro wrote:

On Tue, 16 Jul 2024 10:31:11 +0100, The Natural Philosopher wrote:

On 16/07/2024 01:57, Lawrence D'Oliveiro wrote:

SMART isn’t much use, anyway. I test my storage devices for actual >>>>>>> I/O errors.

That's what you use SMART *for*.

No, I test doing actual I/O.

So does SMART.

No, it extrapolates from its internal firmware behaviour. It tries to
predict failures before they happen.

No it doesn't. It predicts ...

s/predicts/tries to predict/. It’s not a prophet, you know.

And it only catches about 30% of failures.

*plonk*

--
"Corbyn talks about equality, justice, opportunity, health care, peace, community, compassion, investment, security, housing...."
"What kind of person is not interested in those things?"

"Jeremy Corbyn?"
--- SoupGate-Win32 v1.05
* Origin: Agency HUB, Dunedin - New Zealand | Fido<>Usenet Gateway (3:770/3)

T24gMjQvMDcvMjAyNCAwMTozNCwgTGF3cmVuY2UgRCdPbGl2ZWlybyB3cm90ZToNCj4gT24g U3VuLCAyMSBKdWwgMjAyNCAwOTozMTowMyArMDEwMCwgVGhlIE5hdHVyYWwgUGhpbG9zb3Bo ZXIgd3JvdGU6DQo+IA0KPj4gT24gMjEvMDcvMjAyNCAwOTowNiwgTGF3cmVuY2UgRCdPbGl2 ZWlybyB3cm90ZToNCj4+DQo+Pj4gVGhpcyBpcyB3aHkgeW91IGhhdmUgcmVkdW5kYW50IHN5 c3RlbXMuIFRoYXTigJlzIGhvdyB0aGUgcHJvcyBkbyBpdC4NCj4+DQo+PiBOby4gSXRzIHdo eSB5b3UgdXNlIFNNQVJULg0KPiANCj4gQ29tcGFuaWVzIHdob3NlIGJ1c2luZXNzIGl0IGlz IHRvIGVuc3VyZSBkYXRhIGludGVncml0eSBkbyBub3QgcmVseSBvbg0KPiBTTUFSVC4NCg0K Tm8sIHRoZXkgdXNlIGhhcmR3YXJlIFJBSUQgZm9yIHJlZHVuZGFuY3ksIGV4dGVuc2l2ZSBw ZXJmb3JtYW5jZSANCm1vbml0b3JpbmcsIGFuZCByZXRpcmUgbW9zdCBkaXNrcyBiZWZvcmUg dGhleSBmYWlsIGJhc2VkIG9uIHRoZSBzbWFsbCANCnBlcmNlbnRhZ2Ugb2YgZmFpbHVyZXMg b2YgdGhvdXNhbmRzIG9mIG90aGVyIGRpc2NzIG9mIHRoZSBzYW1lIHR5cGUuDQoNCkJ1dCB0 aGF0J3Mgbm90IHdoYXQgdGhlIHR5cGljYWwgcGVyc29uIHdpdGggYSBSYXNwYmVycnkgUGkg YW5kIGEgY291cGxlIA0Kb2YgZGlzY3MgaXMgYWJsZSB0byBkby4gVGhlIFNNQVJUIGluZm9y bWF0aW9uIGdpdmVzIHZhbHVhYmxlIHdhcm5pbmcgb2YgDQpwb3RlbnRpYWwgZmFpbHVyZXMs IHRvIGlnbm9yZSBpdCB3b3VsZCBiZSB0byBlbXBsb3kgdGhlIFNUVVBJRCBmZWF0dXJlIA0K b2YgdGhlIHVzZXIuDQoNCi0tLWRydWNrDQo=
--- SoupGate-Win32 v1.05
* Origin: Agency HUB, Dunedin - New Zealand | Fido<>Usenet Gateway (3:770/3)

On Wed, 24 Jul 2024 21:35:02 +0100, druck wrote:

On 24/07/2024 01:34, Lawrence D'Oliveiro wrote:

Companies whose business it is to ensure data integrity do not rely on
SMART.

No, they use hardware RAID for redundancy, extensive performance
monitoring, and retire most disks before they fail based on the small percentage of failures of thousands of other discs of the same type.

Actually, no. They wait until the disks actually fail before replacing
them.

But that's not what the typical person with a Raspberry Pi and a couple
of discs is able to do. The SMART information gives valuable warning of potential failures, to ignore it would be to employ the STUPID feature
of the user.

Unfortunately, SMART only catches about 30% of potential failures. That’s
why relying on it is not smart.
--- SoupGate-Win32 v1.05
* Origin: Agency HUB, Dunedin - New Zealand | Fido<>Usenet Gateway (3:770/3)

In message <v7s77k$1v8pi$3@dont-email.me>
Lawrence D'Oliveiro <ldo@nz.invalid> wrote:

On Wed, 24 Jul 2024 21:35:02 +0100, druck wrote:

On 24/07/2024 01:34, Lawrence D'Oliveiro wrote:

Companies whose business it is to ensure data integrity do not rely on SMART.

No, they use hardware RAID for redundancy, extensive performance monitoring, and retire most disks before they fail based on the small percentage of failures of thousands of other discs of the same type.

Actually, no. They wait until the disks actually fail before replacing
them.

Anyone with any sense would replace them before the bathtub failure curve starts to rise, which is usually not long after the end of the warranty
period.

But that's not what the typical person with a Raspberry Pi and a couple
of discs is able to do. The SMART information gives valuable warning of potential failures, to ignore it would be to employ the STUPID feature of the user.

Unfortunately, SMART only catches about 30% of potential failures. That's
why relying on it is not smart.

It's smarter than catching 0% of potential failures by waiting until they
have already happened.

David
--- SoupGate-Win32 v1.05
* Origin: Agency HUB, Dunedin - New Zealand | Fido<>Usenet Gateway (3:770/3)

On Thu, 2024-07-25 at 20:20 +0100, David Higton wrote:

Actually, no. They wait until the disks actually fail before
replacing
them.

Anyone with any sense would replace them before the bathtub failure
curve starts to rise, which is usually not long after the end of the
warranty period.

In my previous job, we replaced them when they failed, not before.
Customers don't like paying for hardware that will sit there doing
nothing.
--
Tactical Nuclear Kittens

--- SoupGate-Win32 v1.05
* Origin: Agency HUB, Dunedin - New Zealand | Fido<>Usenet Gateway (3:770/3)

On 25/07/2024 20:20, David Higton wrote:

Anyone with any sense would replace them before the bathtub failure curve starts to rise, which is usually not long after the end of the warranty period.

It entirely depends on usage pattern,. A disk which is being accessed
and having huge amounts of data written and erased, yes.

A disk in a desktop computer that loads the OS and does bugger all.
Many times longer.



--
“Ideas are inherently conservative. They yield not to the attack of
other ideas but to the massive onslaught of circumstance"

- John K Galbraith
--- SoupGate-Win32 v1.05
* Origin: Agency HUB, Dunedin - New Zealand | Fido<>Usenet Gateway (3:770/3)

On Thu, 25 Jul 2024 20:20:38 +0100, David Higton wrote:

Anyone with any sense would replace them before the bathtub failure
curve starts to rise, which is usually not long after the end of the
warranty period.

BackBlaze is a company whose business is data integrity. Every 3 months
they publish a blog post on the reliability stats of the drives that they
use. They don’t replace drives until they fail. Individual failures have essentially zero impact on their business, while replacing drives is a
cost.

Unfortunately, SMART only catches about 30% of potential failures.
That's why relying on it is not smart.

It's smarter than catching 0% of potential failures by waiting until
they have already happened.

The problem is being caught by surprise 70% of the time.
--- SoupGate-Win32 v1.05
* Origin: Agency HUB, Dunedin - New Zealand | Fido<>Usenet Gateway (3:770/3)

On 24/07/2024 21:35, druck wrote:

No, they use hardware RAID for redundancy, extensive performance
monitoring, and retire most disks before they fail based on the small percentage of failures of thousands of other discs of the same type.

This is informative https://www.theregister.com/2024/08/06/backblaze_sees_drive_failure_rates/

---druck
--- SoupGate-Win32 v1.05
* Origin: Agency HUB, Dunedin - New Zealand | Fido<>Usenet Gateway (3:770/3)