Hi

 

Quick summary: I swapped out an old 320GB disk for a 3TB that used to be my parity. I precleared it first, after getting a new parity successfully installed. After adding the 3TB WD Green in place of the old 320GB, it was grinding along doing the data rebuild at about 12MB/s, which is much slower than I had expected, although the preclear took a bit longer than expected too (70hrs)... it was not so long that I thought to investigate further. Anyway, smartctl shows no suspicious numbers at all that I can see, there are no errors in the syslog, and everything would probably have  ended fine (eventually), except that something bizarre happened with my Hamachi connection and I found 1000+ connections to in smbstatus, all from the hamachi ham0 interface. The web UI has seized up, so I can no longer check whether it's even still doing the data rebuild, although unraidd and mdrecoveryd are still using CPU, so it probably is. The worst outcome is that samba/CIFS connections are horribly slow, so slow that I'm surprised when they eventually work at all (1-2 minutes to change folders). This has a bad impact on my WAF, as you might imagine, since the HTPC is effectively crippled without access to any media.

 

Strangely, I can happily navigate around using cd/ls in PuTTY, the shell is as responsive as ever. I tried killing the smbd PIDs associated with the connections coming from the hamachi IP address, most of them died but 4 won't go away even with a kill -9. So chances are I won't be able to shut down the array properly - if I can't stop samba, I won't be able to cleanly unmount. And with no web UI, I can't check the data rebuild to know when it's finished, if it even does!

 

I realise I've gotten myself into this mess, and it's not even that terrible of a mess anyway in that there is no specific risk of data loss or anything, at least not yet and not as far as I can see. But what might happen if I'm forced to do an unclean shutdown and the data rebuild isn't complete? It had progressed to 1TB before the web UI stopped responding, well past the limit of the original 320GB disk, if that matters. (And I still have the original 320GB with all it's data on it, it was too small but not broken or anything.) Will I have to start the data rebuild again? Will unRAID mark the disk bad/red balled? Will it think everything is good but leave me with invisibly bad parity that I won't notice until I next have to do a data rebuild?

 

Anyway, I'd really appreciate any opinions on what actions you think I could/should take.

 

Oh, pertinent details. Running 5.0 but with a custom kernel (you can roll your eyes now... *custom kernel*!! I know, I know! In my defense, it's been running fine for almost a year The 3TB disk is a WD Green, model WD30EZRX-00DC0B0, and there are 13 disks plus cache plus parity in the array, with parity at 4TB. I have attached smartctl output for the disk being rebuilt (/dev/sdf aka disk9), which certainly seems not to have any obvious bad things going on. I've also included a partial syslog - I know it's preferable to send the whole thing but it's big due to mover logs, and I'd just rather not unless it seems really worth it. I have kept a copy of it though, and the array is fully accessible via PuTTY so unless things change for the worse, I can do further diagnostics as necessary.

 

Thanks,

-- tallorder

syslog-brief.txt

smartctl-sdf.txt

Ah, if anyone is interested, turned out to be nothing to do with the weird 1000's of samba connections. Rather, my Dropbox plugin had gone nuts and was blocking lots of things. When I eventually killed it (after my slow-but-successful data rebuild had seemingly completed and mdrecoveryd had gone idle), my whole system went back to normal and although the web UI never returned, I was able to do a clean shutdown. After restart, all good.

 

-- tallorder