Full format new harddisks?

[bononos]

Do you format (not the quick format) the entire harddisk before using it? I haven't done it for a long time because it takes such a long time. Now after running into another bad sector problem again, I'm wondering if I should've done a full disk scan/format first.

 

[Ketchup]

Nope. Haven't done a full format in a decade at least (on a Microsoft system anyway).

 

[corkyg]

I have always done that initially in order to get a good map of the drive. Yes, it takes time, but that can be managed.

 

[VeryCharBroiled]

i do a full format on new drives. yeah it take about three years. makes me feel better even its not need.

 

[Elixer]

I also do a full format, I have found flaky drives that way.

 

[bononos]

I did some searching for some sort of scanning utility and found hddscan3.3 (by Artem Rubtsov) useful because it runs on Windows and the start/end endpoints can be specified for read/write tests so large drives can be worked on over several sessions and it can generate a list of slow sectors.

Both of the WD drives which failed gave smart errors, I caught the 2nd one early on before I lost anything.

 

[mikeymikec]

Nope. Haven't done a full format in a decade at least (on a Microsoft system anyway).

Ditto. I've had a grand total of one DOA HDD since 2003 AFAIK, and that was in the era that I did full formats of new drives. The issue became apparent during Windows setup, after the format.

I do however do a full chkdsk (/f /v /r) of every drive I purchase.

 

[ronbo613]

Most of the time I do a full format even if it is not required. I look at it as the initial shakedown test of the drive. Those large capacity drives can take some time but it helps keep the house warm in the winter.

 

[Phynaz]

Nope. Haven't done a full format in a decade at least (on a Microsoft system anyway).

Same here.

 

[BonzaiDuck]

I stick to old habits established before 1995. I just want to assure myself that a drive will format flawlessly before I move on with it. I always run a diagnostic program against new large capacity HDDs. I check to make sure new SSDs are properly aligned.

 

[fastman]

When installing OS, I do the default format, quick I guess and never have had a problem.

 

[PliotronX]

No way, I am too impatient. It was fine when 6GB was the largest drive but to full format TBs? No thanks. Same for creation of fixed VHDs, I'll use vhdtool or vhdxtool.

 

[ch33zw1z]

Sure do. Nothing to lose.

I also have multiple backups of data, so no need to rush.

Think that takes a while, wait until you prep your LUKS storage disk afterwards...

 

[corkyg]

LUKS = Linux Unified Key Setup.

 

[CiPHER]

I don't really see the point of formatting a new harddrive. If the purpose is to find disks that die rapidly, then you should read not write since things like bad sectors will not get detected upon writing.

The only real valid point to format a harddrive is when using LUKS or other form of full-disk encryption. Before you apply the encryption, it is best of the harddrive has all random data written to it, and not all zeroes so that it gives information which parts of the harddrive were written with encrypted data.

For 'stress testing' harddrives I suggest using ZFS since ZFS detects corruption as well. So leaving it on for a night with heavy I/O and scrubbing will tell you if the drive made a mistake. You cannot determine that if all you do is formatting.

I have always done that initially in order to get a good map of the drive. Yes, it takes time, but that can be managed.

What do you mean with 'get a good map of the drive' ?

 

[corkyg]

I don't really see the point of formatting a new harddrive. If the purpose is to find disks that die rapidly, then you should read not write since things like bad sectors will not get detected upon writing.

The only real valid point to format a harddrive is when using LUKS or other form of full-disk encryption. Before you apply the encryption, it is best of the harddrive has all random data written to it, and not all zeroes so that it gives information which parts of the harddrive were written with encrypted data.

For 'stress testing' harddrives I suggest using ZFS since ZFS detects corruption as well. So leaving it on for a night with heavy I/O and scrubbing will tell you if the drive made a mistake. You cannot determine that if all you do is formatting.


What do you mean with 'get a good map of the drive' ?

Full format creates a fresh map of the drive where bad sectors are hidden, etc., so they can't be used.

 

[CiPHER]

Full format creates a fresh map of the drive where bad sectors are hidden, etc., so they can't be used.

Well, a bad sector first has be known. If a sector is 'bad' but the harddrive does not know it - meaning it never failed a READ request - then overwriting that sector would not uncover the issue. Even worse, it would destroy any evidence of there ever been a bad sector.

If a bad sector is known, meaning it was at least read once and failed, then the harddrive will show Current Pending Sector > 0 in the SMART-output. In this case, if you overwrite the bad sector, the harddrive will actively READ that same sector afterwards, and check whether it is actually readable. If not, it will replace the sector with a reserve sector. But it will not do this test if the harddrive is unaware of the weak/bad sector.

Additionally, old filesystems like FAT and NTFS have a special file $BadSect$ which stores sectors that the filesystem should not use. This is extremely outdated, since for many decades harddrive replace bad sectors themselves. The filesystem should not mimic this task, which is redundant. Newer filesystems have no such feature, but are designed to cope with individually unreadable sectors (ZFS, Btrfs, ReFS).

So for a new harddrive, reading the entire surface actually uncovers bad sectors, but simply writing does not. Even worse, after writing, unreadable sectors may become readable again - so called uBER bad sectors that are only unreadable due to insufficient error correction and not because of physical damage. After (over)writing these, all evidence disappears from the SMART data; Current Pending Sector is substracted but Reallocated Sector Count stays the same.

Stress testing with some kind of checksum utility, like the ZFS filesystem, would have my preference to validate new harddrives.