On Tuesday, November 10th, 2026 at 15:46, Tom Smyth <[email protected]> wrote:
> we have had servers running for 5+ years with SSDs that never failed > with 10% free space... > I think in the 7th year ... I have lost 2 drives in that setup > > we have had 20+ devices running windows without that precaution fail > within 12 months without that precaution... > I suppose the question is what is your risk appetite... > SSDs wear out ... TLCs etc... > > it is not entirely scientific but it is an indicator and alot of > different people recommend it and > alot of "enterprise drives" have overprovisioned space to account for wear .. > > 1Tb to the user and more under neath not exposed to the user... > > > > On Tue, 10 Nov 2026 at 21:39, Brian Brombacher <[email protected]> wrote: > > > > > > > On Nov 10, 2026, at 4:16 PM, Roderick <[email protected]> wrote: > > > > > > Thanks, Tom. > > > > > > I suppose you calculate as in the web page of Kingston: > > > > > > Percentage Over-provisioning = (PhysicalCapacity - UserCapacity) / > > > UserCapacity > > > > > > Hence, 50% means 1/3 of the whole capacity for Overprovisioning. > > > > > >> Am Di., 10. Nov. 2026 um 20:42 Uhr schrieb Jan Stary <[email protected]>: > > >> > > >> Surely the controller knows nothing about "partitions". > > > > > > But know about unused blocks, and this will never be used if it is in a > > > unused > > > partition. > > > > > > I marked the whole disk as unused with FreeBSD command: > > > > > > camcontrol security adaX -U user -s pass -e pass # literally > > > (sustitute only X with device nr) > > > > > > If disc is frozen (seen with: camcontrol security adaX), hot unplug and > > > plug data cable after 10 seconds. > > > > > > I hope that is OK, and it would be interesting to know if can be done with > > > OpenBSD. > > > > > > > I’m curious how many of you have actually experienced a disk failure due to > > not having enough spare blocks for SSDs manufactured after 2020. I write a > > LOT of data to SSDs and I have yet to experience this issue and I take no > > precautions such as saving space at the end of the disk with an unused > > partition. Now that I just said that, I’ll have 10 disks fail tonight, > > karma… When I have a SSD drive on the computer, I try to keep only files on the SSD that rarely or never change in order to limit disk writes. So I have a conventional hard drive for many directories. Also, my /tmp directories are always memory resident in OpenBSD and Linux. I use the /tmp directory for far more than just what it is used for by default. For example, if I'm creating a tar file to be sent to the cloud, I create it in /tmp, transmit it to the cloud, and then remove it from /tmp. And when compiling code, the obj files are stored in a directory in /tmp such as /tmp/billyjack/obj. My /var/log may be on either the SSD or a conventional drive, but instead of creating a lot of churn by rotating them often, I prefer to rotate them no more often than once every three months. Once a year is better. The /var/log/authlog file would get enormous if it weren't for the fact that access to ssh on my servers are all limited to a whitelist of sources. For times when we don't know what IP addresses may be used to connect, we whitelist some of a VPN provider's endpoints. That helps limit the size of the log files. Also, should there be a zero day found in the ssh daemon, the number of attackers testing it should be very limited since they would have to be using the same VPN and through the same cities as we have whitelisted. In the several months since we switched to this, we have seen no unauthorized attempts to connect to ssh on any of our servers. By the way, /var/log/authlog is flagged sappend so that if an attacker actually did get in (rather unlikely considering the above limits and the requirement to use three distinct ssh keys -- no passwords allowed), he would find it much more difficult to try to cover his tracks. The first step to rotating it is to change to single user mode. Billy Jack
signature.asc
Description: OpenPGP digital signature

