Re: GPT boot with ZFS RAIDZ "ZFS: i/o error - all blockcopies unavailable"

From: Paul Wootton <paul_at_fletchermoorland.co.uk>
Date: Fri, 26 Jun 2009 19:54:28 +0100
Steven Hartland wrote:
>
> ----- Original Message ----- From: "Paul Wootton" 
> <paul_at_fletchermoorland.co.uk>
>
>> As bad as it sounds, YAY, im not the only one...
>> If only I had seen this thread earlier - As of 1 hours ago im the new 
>> owner of another 500Gb drive. I had already checked the SMART data on 
>> the drives and it showed everything was good, but decided some new 
>> drivers (getting one at a time) would help me eliminate the drives as 
>> being the source of the LBA errors.
>>
>> Interesting though, my LBA errors are also around the 429xxxxxxx 
>> number, usually with the 3rd of 4th number being somewhere around 8xxx.
>> After rebuilding and populating the pack many many times with various 
>> versions of 8 I did manage to get a pack that would try and boot the 
>> kernel with out any LBA or IO errors but I had forgotten to add the 
>> zpool.cache file. I rebooted on a working drive/installation, pushed 
>> the zpool.cache file across and tried booting off raidz hoping that 
>> all would be good and happy, but no, the LBA errors started appearing...
>
> You might want to check what LBA mode is being used. Some disks require
> you to use 48bit LBA for all of the disk if supported and I've seen
> some drivers use it only if they "think" its required for the position
> in question, which can cause issues with the firmware on some drives,
> resulting in LBA errors.
>
>    Regards
>    Steve
>
Im unsure how to check what LBA mode my drives are currently using, but 
I have done some more testing with interesting results.

I am using 3 drives (ad18, ad20 and ad24) and all done with GPT tables.

Made a pack and have a 1gb zfs partition on each drive - raidz boots 
fine and the machine runs happily
Increased the partition on ad24 to the drive's maximum capacity - 75G. 
Rebooted and the machine is still happy.
Increased the partition on ad20 to 3.8G. Rebooted and the machine is 
still happy
Increased the partition on ad18 to 3.8G. All is still happy.
Increased the partition on ad20 to 7.6G. Again, all happy
Increased the partition on ad18 to 7.6G. I now start getting the LBA errors

I still need to try ad18p2 at 3.8G, ad20p2 at 7.6G, ad24p2 at 75G and 
try booting off ad20 instead of ad18

During other tests, after increasing the parition size, the kernel would 
boot but fail to load the root filing system. I pointed it to a UFS 
based root to bring the system online. I took a look at the zpool status 
and it shows 1 of the partitions as corrupt and the whole pack as faulty 
and not degraded.

Does anyone else see this sort of behavior?

Paul


-----------------------------------------------------------------------------------
Fletcher Moorland Limited is a company registered in England and Wales. 
Registration number: 2984467. 
Registered office: Elenora Street, Stoke on Trent, Staffordshire, ST4 1QG. 
VAT Registration number: 478730606 
Telephone: 01782 411021 | Fax: 01782 744470 | http://www.fletchermoorland.co.uk 
Received on Fri Jun 26 2009 - 16:55:16 UTC

This archive was generated by hypermail 2.4.0 : Wed May 19 2021 - 11:39:50 UTC