Score:0

How to Diagnose Constant ZFS Repairs

ru flag

I ran a zpool scrub 4 times on my pool today. Each time, it found and fixed errors. Bad drive, right?

The pool has 2 drives, and smartctl reports zero Raw_Read_Error_Rate for each. So the drives say nothing is wrong. ZFS is finding errors on both drives.

The drives are external M.2 SATA drives in M.2-to-USB adapters, connected by 6-inch cables that came with the adapters. The PC is an HP EliteDesk mini.

It didn't do this a month ago. This is the first time I've seen ZFS errors in the 6 months since I set this PC up. I've not recently updated Linux. Since smartctl reports no read errors, I conclude that the error must me in the:

  • M.2-to-USB adapters
  • cables
  • USB ports
  • motherboard

Is there a better way to go about diagnosis than replacing the components one at a time, starting with the cables, since they're cheapest?

What would make both drives begin to show errors in ZFS but not on the drives? It seems wildly improbable that both adapters/cables/USB ports would begin failing at the same time.

Score:0
ca flag

External USB adapters can be quite flaky when reading or writing vast amount of data. You can buy a cheap internal PCI-E to M.2 SATA card to rule out any issue due to the external USB adapters.

Kevin Kleinfelter avatar
ru flag
Good thought. An HP EliteDesk is similar form factor as a Mac Mini, with no slots. I'm moving into a small apartment, so I replaced my full-sized tower with a small form factor EliteDesk, so those drives are going to have to remain external. Replacing an adapter is second on my list. If there are known-good external adapters, please suggest!
shodanshok avatar
ca flag
@KevinKleinfelter Well, I don't have much trust in external adapters. That said, Sabrent adapters seem good to me (but this is highly debatable).
Kevin Kleinfelter avatar
ru flag
I ran a dozen iterations of memtest86 without error; it probably isn't RAM. Then I reseated the drives, the USB cables, and the power supply cable. It ran error free. Ran it again and got errors. The drive enclosures are hot to the touch. I'm going to try scrub/pause/scrub/pause... and see if they run cooler+cleaner. I'll also try one with the enclosure removed. I ordered a Sabrent adapter last night (based on the reviews). When it arrives, I'll swap a drive into it and if it is better. I'll also try putting a fan on them.
Kevin Kleinfelter avatar
ru flag
Purchased a Sabrent adapter. Replaced one of the DMLIANKE adapters. The drive with the Sabrent adapter no longer has errors. The remaining drive with the DMLIANKE adapter has errors -- solely under heavy lead. Clearly the DMLIANKE adapters can't handle heavy load. Ironically, they corrupt data only when I'm running a ZFS scrub!
mangohost

Post an answer

Most people don’t grasp that asking a lot of questions unlocks learning and improves interpersonal bonding. In Alison’s studies, for example, though people could accurately recall how many questions had been asked in their conversations, they didn’t intuit the link between questions and liking. Across four studies, in which participants were engaged in conversations themselves or read transcripts of others’ conversations, people tended not to realize that question asking would influence—or had influenced—the level of amity between the conversationalists.