Jim Klimov
2013-12-21 15:13:07 UTC
Hello all,
I got access to my old Home-NAS again, which got me looking under deep
the hood of ZFS in the first place due to strange errors with my pool.
This is a Pentium-4 based PC including an Asus P5B-Deluxe motherboard
with 7 SATA connectors (an Intel and a JMicron set), with 8GB of
(non-ECC) memory. Used to be quite a machine back in its day as a PC
and gaming station! ;)
Currently it serves as an OpenIndiana-based storage unit, but serves
poorly: despite using raidz2 over 6*2Tb drives, it keeps finding errors
(and DD'ing the offsets from disks shows that indeed there is trash on
disk instead of proper data). Also the system disk finds and fixes some
checksum errors on every scrub - luckily, it uses copies=2. While there
was only the old 80Gb SATA (which I thought had died - but did not) it
usually found 2-6 errors per scrub. Now I have mirrored it with a newer
250Gb SATA (picked from an HP Microserver barebone) to migrate the OS
from an old disk, and it finds and fixes up to 30 errors per scrub.
My guess is that these problems may be due to randomness from overheat
in the CPU or chipset - but no real complaints here, insufficient power
somehow or lack of ECC. Or just plain age is showing... All I can say
instrumentally is that long SMART tests did not find any errors.
So far I am trying to evacuate the remaining data from this box to
some other storage. But I am not sure what to do with it or its parts
such as disks - can they be trusted to rebuild a new pool, for example?
Or should they be safer put away or repurposed back into a PC?..
Any ideas or comments?
Thanks,
//Jim
I got access to my old Home-NAS again, which got me looking under deep
the hood of ZFS in the first place due to strange errors with my pool.
This is a Pentium-4 based PC including an Asus P5B-Deluxe motherboard
with 7 SATA connectors (an Intel and a JMicron set), with 8GB of
(non-ECC) memory. Used to be quite a machine back in its day as a PC
and gaming station! ;)
Currently it serves as an OpenIndiana-based storage unit, but serves
poorly: despite using raidz2 over 6*2Tb drives, it keeps finding errors
(and DD'ing the offsets from disks shows that indeed there is trash on
disk instead of proper data). Also the system disk finds and fixes some
checksum errors on every scrub - luckily, it uses copies=2. While there
was only the old 80Gb SATA (which I thought had died - but did not) it
usually found 2-6 errors per scrub. Now I have mirrored it with a newer
250Gb SATA (picked from an HP Microserver barebone) to migrate the OS
from an old disk, and it finds and fixes up to 30 errors per scrub.
My guess is that these problems may be due to randomness from overheat
in the CPU or chipset - but no real complaints here, insufficient power
somehow or lack of ECC. Or just plain age is showing... All I can say
instrumentally is that long SMART tests did not find any errors.
So far I am trying to evacuate the remaining data from this box to
some other storage. But I am not sure what to do with it or its parts
such as disks - can they be trusted to rebuild a new pool, for example?
Or should they be safer put away or repurposed back into a PC?..
Any ideas or comments?
Thanks,
//Jim
