Very slow disk replacement/resilver (RAIDZ2)

[deafen]

As part of a general refresh of a bunch of stuff, I put in an LSI SAS2008 card (the Dell variant, with IT firmware). I'm sure I'll have another failure at some point, and I'll revisit this then.

 

[deafen]

... aaaaaand, I've had another drive start throwing uncorrectable sectors - this time, one of the replacements that Seagate sent me! And it's beyond the pitiful replacement warranty period. So I have an HGST NAS-rated drive on order to swap in, and I'll get to see if my resilver performance is any better with the LSI SAS2008 card I've put in. But to confound the results, I also increased the size of the pool to 7 disks from 6, so I can't isolate any single variable if performance does improve. Eh, oh well.

I will say that my scrub speed with the new setup is maxing out at 582 MB/s, so hopefully only good things now.

 

[cyberjock]

I hope you don't mean that you added a single disk to the zpool. If you did you just broke redundancy for your pool. :(

 

[deafen]

Oh no, that's exactly what I did! Because I don't know anything! Thank you, oh mighty savior!

Code:

[root@delta ~]# zpool status
  pool: tank
state: ONLINE
  scan: scrub in progress since Tue Sep  9 12:58:56 2014
        1.35T scanned out of 9.54T at 540M/s, 4h24m to go
        0 repaired, 14.19% done
config:

        NAME                                            STATE     READ WRITE CKSUM
        tank                                            ONLINE       0     0     0
          raidz2-0                                      ONLINE       0     0     0
            gptid/58db802d-30ab-11e4-b062-6805ca277778  ONLINE       0     0     0
            gptid/5970ba66-30ab-11e4-b062-6805ca277778  ONLINE       0     0     0
            gptid/5a00aacd-30ab-11e4-b062-6805ca277778  ONLINE       0     0     0
            gptid/5a90b41f-30ab-11e4-b062-6805ca277778  ONLINE       0     0     0
            gptid/5b445f85-30ab-11e4-b062-6805ca277778  ONLINE       0     0     0
            gptid/5c0ccd1a-30ab-11e4-b062-6805ca277778  ONLINE       0     0     0
            gptid/5cde6e1e-30ab-11e4-b062-6805ca277778  ONLINE       0     0     0

errors: No known data errors
[root@delta ~]#

/sheesh

 

[danb35]

If you'd seen how many times in recent history we've seen someone do exactly that, maybe you wouldn't be such a smartass with your post.

 

[deafen]

Sorry, just tired of the condescending attitude. He's already refused to actually help me at all with this issue BECAUSE AMD, and now he's throwing something irrelevant out there that assumes I have no idea what I'm doing. Pretty tiresome.

 

[Ericloewe]

Sorry, just tired of the condescending attitude. He's already refused to actually help me at all with this issue BECAUSE AMD, and now he's throwing something irrelevant out there that assumes I have no idea what I'm doing. Pretty tiresome.

I'm sorry, but it's a valid question, given the way your reply was phrased. It was my first thought, too.

Nobody here means any offence (unless you really piss someone off), it's just good to follow up on little details that many miss.

 

[cyberjock]

Sorry, just tired of the condescending attitude. He's already refused to actually help me at all with this issue BECAUSE AMD, and now he's throwing something irrelevant out there that assumes I have no idea what I'm doing. Pretty tiresome.

It wasn't irrelevant. And if you want to be pedantic about it I could easily throw you a warning for that. I've been told I need to not let users be rude "because they've decided they don't like something/someone".

My question was completely valid and we've had tons of people that have done *exactly* that. I was trying to help you, but if you want to be a jerk I can be one too...

Maybe you should take what I said to heart and realize that even if I hate your guts I'm still looking out for your data. But if you are too busy with the attitude there's a simple "ban user" button that solves that problem.

 

[deafen]

Eh, maybe that got a little personal, and for that I apologize. The validity of your question isn't at issue, just its relevance - given that my question is about resilver performance, not redundancy compromise - but I guess I'll just have to agree to disagree.

 

[deafen]

Now resilvering at a blistering 575 MB/s scanning speed. So my original problem was either the result of the drive model (unlikely, but possible) or a crappy SATA chipset (way more likely). CPU is the same, so it's not an AMD issue per se, but this board does have a pretty low-end southbridge (760G).

Pretty happy with that $80 LSI card purchase now!

 

[gpsguy]

I'm glad to hear that things are humming away now. One other change you made, was the removal of the cache device. That probably helped a bit.