On 07/29/2015 08:34 AM, Volker Armin Hemmann wrote: > Am 29.07.2015 um 16:52 schrieb Daniel Frey: >> On 07/28/2015 12:04 AM, Volker Armin Hemmann wrote: >>> you know - this does not sound like ssd failure. Most SSDs bomb out by >>> just becoming completely unacessible. >>> >>> dmesg errors? >> Filled with /dev/sda errors when it "failed". > > oh the joy.
Yeah. I don't remember what the exact error message was, other than it was filled with "can't read" and "can't write" in the messages. Booting from USB worked fine. Compiling while booted from USB worked until I chroot'ed to the failed SSD. Then a bunch of segfaults and other weird errors. I've seen this before both at home and work, where SSDs do this with no warning. IMO they're way too unreliable. I don't have one in my server or workstation at home. I had one in my server and it lasted two years before similar issues above. It (server SSD) was replaced with a spinning disk 3 years ago and it's still running today. I'm pretty sure if I replaced it with another SSD it would've failed already. > >> >>> are you using ecc ram? >> Nope. >> >>> if not - maybe, just maybe it is your ram at fault. The stuff the kernel >>> sends and the stuff that end on the ssd might not be identical. >>> >> Ran memtest overnight on it, no errors. >> >> Dan >> >> >> > > I had ram that passed memtest - and zfs detected errors. Went ecc ram, > no more errors. > > With ram hammer as latest attack vector, ecc is even more worth its money. > ECC is fine and dandy, but this motherboard doesn't support it. It's a desktop board from 2008. All it does is a frontend for mythtv, nothing else. Dan
