Raspberry Pi 2B and SSD drive

Karl Denninger karl at denninger.net
Sun Jun 17 19:04:08 UTC 2018

On 6/17/2018 13:55, Per olof Ljungmark wrote:
> Hi,
> Quoting bob prohaska <fbsd at www.zefox.net>:
>> On Sun, Jun 17, 2018 at 03:21:33PM +0200, Per olof Ljungmark wrote:
>>> FBSD 11.2-RC1
>>> How can I make good use of a USB SSD drive connected to the Pi?
>> It might help if you indicated what constitutes "good use" 8-)
> As just stated, a light duty smtp-server, and I need better
> reliability, not speed, than what a SD card usually offers.
>>> For instance, can I start from the SD card and then hand over
>>> everything
>>> to the SSD drive? From googling it does look like it is possible.
>> Going out on a limb here, but I don't think it'll make much difference.
>> Personally I like to leave root on microSD and keep /usr, /var, /tmp
>> and swap on USB flash. Far as I know USB is the only mass storage
>> connection interface, so that bottleneck is unavoidable.
>> Having the booting bits separate from the scribbling bits offers a
>> little
>> protection against a totally unbootable system in case of user error or
>> other mishap. It's still necessary to make provision for a serial
>> console,
>> since last I checked one can't talk to single user via USB keyboard.
> Makes sense of course.
IMHO do not put "scribbling bits" on SD card; they're not made for it,
it's wildly abusive to them to do so, and they *will* fail if subjected
to that in a moderate-to-heavy load environment, at which point the unit
usually panics and is unbootable when it comes back up.

Oh, and it's also wildly *slow* too, so there's another reason not to.

Putting that on a USB-attached spinning rust device works reasonable
well; an SSD is gross overkill for this application and likely not much
help too, as an SSD that takes an unexpected power loss and is not
designed to be power-fail tolerant has a decent chance of coming back
un-fsckable (or worse, it fsck's ok but has corruption that is invisible
on it that causes an immediate panic when referenced) due to the lack of
coherence in writes when the power goes off.

Power-fail-safe SSDs are expensive; the only one I've found that's not
ridiculously-so are the Intel 730 series, of which I own several -- they
have passed repeated "plug pull" tests while running a Postgres test app
that is intentionally designed to detect any sort of game-playing when
it comes to claims of "I committed that write" when it really did not.

Karl Denninger
karl at denninger.net <mailto:karl at denninger.net>
/The Market Ticker/
/[S/MIME encrypted email preferred]/
-------------- next part --------------
A non-text attachment was scrubbed...
Name: smime.p7s
Type: application/pkcs7-signature
Size: 4897 bytes
Desc: S/MIME Cryptographic Signature
URL: <http://lists.freebsd.org/pipermail/freebsd-arm/attachments/20180617/39cdd7f0/attachment.bin>

More information about the freebsd-arm mailing list