ZFS RAIDz Issue and Labels

[fayraz]

Hi,

I have been testing ZFS RAIDZ with 3 disks but when I unplug one disk the whole array goes down?

I have 3 disks: 2 x Seagate and 1 x Hitachi all 1TB. When I unplug the Seagates either one of them the array goes down but when I unplug the Hitachi the array is degraded and I can still access the data.

I am using Ubuntu Server and installed ZFS-FUSE.

The other issue I am having is I cannot rename labels on the volumes, I always get a superblock error; I have e2label?

 

[parityboy]

@OP

Not familiar with ZFS-FUSE, but I'd imagine that e2label wouldn't work on ZFS - I'd guess that the superblocks are in the wrong place, or that the FUSE handler for ZFS is incomplete. Could be wrong though. Have you looked at zfsonlinux?

 

[fayraz]

I will try using zfsonlinux, does anyone know how I can remove ZFS-FUSE, I don't want it to conflict with zfsonlinux?

Thanks.

 

[danswartz]

I assume with 'aptitude purge zfs-fuse' or somesuch?

 

[fayraz]

I assume with 'aptitude purge zfs-fuse' or somesuch?

Excellent, that worked. I'll see how I can get zfsonlinux installed.

 

[fayraz]

I don't understand whats happening when I try to use ZFS I have been using zfsonlinux and created a mirror of two identical disks and there is no redundancy when I pull one out. The same thing happens when I create a RAIDZ array, below is the output I get when pulling one disk out of the mirror:

pool: tank2
state: UNAVAIL
status: One or more devices could not be used because the label is missing
or invalid. There are insufficient replicas for the pool to continue
functioning.
action: Destroy and re-create the pool from
a backup source.
see: http://zfsonlinux.org/msg/ZFS-8000-5E
scan: none requested
config:

NAME STATE READ WRITE CKSUM
tank2 UNAVAIL 0 0 0 insufficient replicas
mirror-0 UNAVAIL 0 0 0 insufficient replicas
sda FAULTED 0 0 0 corrupted data
sdb FAULTED 0 0 0 corrupted data

 

[mwroobel]

I don't understand whats happening when I try to use ZFS I have been using zfsonlinux and created a mirror of two identical disks and there is no redundancy when I pull one out. The same thing happens when I create a RAIDZ array, below is the output I get when pulling one disk out of the mirror:

pool: tank2
state: UNAVAIL
status: One or more devices could not be used because the label is missing
or invalid. There are insufficient replicas for the pool to continue
functioning.
action: Destroy and re-create the pool from
a backup source.
see: http://zfsonlinux.org/msg/ZFS-8000-5E
scan: none requested
config:

NAME STATE READ WRITE CKSUM
tank2 UNAVAIL 0 0 0 insufficient replicas
mirror-0 UNAVAIL 0 0 0 insufficient replicas
sda FAULTED 0 0 0 corrupted data
sdb FAULTED 0 0 0 corrupted data

What drives are you using and how are they connected? What controller card are you using and how is it connected to the drives? Directly cabled, backplane? If you have a working mirror, shutdown instead of pulling out, pull out while the machine is off and then bring it back up does tank2 stay up with the broken mirror?

 

[fayraz]

What drives are you using and how are they connected? What controller card are you using and how is it connected to the drives? Directly cabled, backplane? If you have a working mirror, shutdown instead of pulling out, pull out while the machine is off and then bring it back up does tank2 stay up with the broken mirror?

I am using a HP Microserver, got the OS running off the ODD port so got 4 bays free. They are all 1TB, 2 x Seagates and 1 x Hitachi.

When I pull the drive its always with the machine off and I thought the RAIDz was failing because I had the Hitachi drive so I thought I'd create a mirror with just the Seagates and the same thing happens.

 

[staticlag]

I'd test the backplane first, this sounds like a hardware error.

 

[fayraz]

I'd test the backplane first, this sounds like a hardware error.

The hard drives are being detected so I presume the backplane is OK or is this not the case? Is there any other way to check the backplane?

It is a strange issue because when I was using RAIDz with 3 disks and removed the Hitachi drive the array stayed up and said it was degraded its just when I unplug any of the Seagates.

 

[staticlag]

try re-arranging the drives in the array, by trial and error you can figure out if one port is the culprit

 

[bexamous]

Pretty sure I've had this problem, don't use '/dev/sda' when creating array... use 'dev/disk/by-*/something' instead. Eg with mirror, you create it and seagate is sda and hitatchi is sdb, you power off unplug seagate, hitatchi is now sda, and it says sda is corrupted and sdb is missing.

Another fix, is export the messed up pool, and then just import it. When importing it won't be confused that 'sda' is now what 'sdb' used to be and now your array will work again.

Also BTW, you should be sure test with pulling a drive while system is both off and on, both are important, if anything while on is more important.

 

[bexamous]

See here: http://zfsonlinux.org/faq.html#WhatDevNamesShouldIUseWhenCreatingMyPool

 

[fayraz]

Pretty sure I've had this problem, don't use '/dev/sda' when creating array... use 'dev/disk/by-*/something' instead. Eg with mirror, you create it and seagate is sda and hitatchi is sdb, you power off unplug seagate, hitatchi is now sda, and it says sda is corrupted and sdb is missing.

Another fix, is export the messed up pool, and then just import it. When importing it won't be confused that 'sda' is now what 'sdb' used to be and now your array will work again.

Also BTW, you should be sure test with pulling a drive while system is both off and on, both are important, if anything while on is more important.

See here: http://zfsonlinux.org/faq.html#WhatDevNamesShouldIUseWhenCreatingMyPool

Thanks for the info, I will give this a go.

 

[fayraz]

Pretty sure I've had this problem, don't use '/dev/sda' when creating array... use 'dev/disk/by-*/something' instead. Eg with mirror, you create it and seagate is sda and hitatchi is sdb, you power off unplug seagate, hitatchi is now sda, and it says sda is corrupted and sdb is missing.

Another fix, is export the messed up pool, and then just import it. When importing it won't be confused that 'sda' is now what 'sdb' used to be and now your array will work again.

Also BTW, you should be sure test with pulling a drive while system is both off and on, both are important, if anything while on is more important.

I just tried using the full device IDs and created a RAIDz array and it worked! Thanks for the info.