On Oct 25, 2026, at 9:50 pm, bob prohaska <[email protected]> wrote:
> On Fri, Oct 23, 2026 at 06:27:43PM -0700, bob prohaska wrote: >> >> The least-intrusive change I can make at this point is replace the >> wallwart on the powered hub running the hard disk. That will change >> only one connection, the barrel jack on the hub. Wiggling it once >> raised the measured voltage on the hub by about 100 mV, so that >> disturbance is already taken into account. > > Turns out the hub barrel jack is ever-so-slightly smaller than that > on the spare wallwart.... Meanwhile, no hint of faults emerged > while watching the original power supply in action with a 'scope. > >> The machine is still running buildworld at the building libraries >> stage. I'll wait for it to finish or crash before changing things. > > Buildworld stopped again, same sort of filesystem errors. It looks > like time to test the disk, but it isn't smartmontools compatible. > > Can anybody suggest a disk testing utility that's non-destructive > and usable on either FreeBSD or RasPiOS? I've already compiled > disroaster, but it's obligately destructive, so not ideal. Just seeing whether the drive generates errors when reading can be instructive because it suggests the health of the drive is not good. You can try to read the whole drive just using the builtin FreeBSD "dd" command. Alternatively, you can use some of the versions in FreeBSD Ports that are slanted towards digital forensics acquisition, like sysutils/dc3dd, sysutils/dcfldd, sysutils/dd_rescue, and sysutils/ddrescue. (These focus more on skipping past errors, or trying especially hard to read blocks with errors.) If you can attach the drive to a controller that lets you read the S.M.A.R.T. data, that could be instructive. Values of Reallocated_Sector_Ct, Current_Pending_Sector, and Offline_Uncorrectable attributes can be illuminating in giving an idea of the overall health of the drive. A non-zero value for Reallocated_Sector_Ct would indicate bad sectors on the drive that have been successfully reallocated. IIRC, a non-zero value for Current_Pending_Sector would indicate an ongoing attempt to reallocate a bad sector that could point towards the drive being out of spare sectors with which to reallocate. > I'm trying a linux program called diskscan, but it just passively > reads the disk. A read/write/read/restore strategy would be more > informative, even if more risky. A read/write/restore strategy is very risky. If the drive is failing, this could lead to more corruption if the "restore" segment of the triad of operations fails. IMHO, the only way to test the write health of a suspect drive that is in service is to copy it to a replacement drive and then remove the source drive from service and test it separately. Even read-testing a drive can impose a load on it that could provoke it to fail. Cheers, Paul.
