moving multiple drives from RFS to XFS

[dikkiedirk]

I'm updating my test server and freed up a slot in the case where I am gonna mount a freshly precleared disk. The disks in this server are all RFS so I am gonna take this opportunity to switch them to XFS. Can I use MC to copy over the files in IPMI or is there a better way like Rsync? Should I disable parity to speed up things or can I preserver parity? Any advice is welcome. Running unraid 6.3.2 BTW.

Edited March 1, 2017 by dikkiedirk

[RobJ]

Please see File System Conversion

[dikkiedirk]

Thanks. Rsync does look easier than mc. Never used it though. Many prerequisites but that guide looks fairly straight forward. Waiting till the parity check ends and then let's go. Can it be done in IPMI?

[RobJ]

10 hours ago, dikkiedirk said:

Can it be done in IPMI?

 

If it gives you a standard terminal console.

[dikkiedirk]

How long could converting a 2 TB disk take? After about 8 hours it has copied 900 MB. It are not the fastest drives connected to a M1015 controller on an Asrock motherboar with E3-1230 V3.

[RobJ]

1 hour ago, dikkiedirk said:

How long could converting a 2 TB disk take? After about 8 hours it has copied 900 MB. It are not the fastest drives connected to a M1015 controller on an Asrock motherboar with E3-1230 V3.

 

Not positive, but that does seem slow.  I'd check the SMART reports when you have a chance.  And when the copy is done, you might want to try the Drive Performance tester, diskspeed.sh, see if any drives are unusually slower than they should be.  Also, take a quick look through the syslog for drive errors.  If they are impacting your speed, they should be obvious.

[trurl]

2 hours ago, dikkiedirk said:

After about 8 hours it has copied 900 MB

Did you really mean 900GB? 900MB in 8 hours sounds like a bad dialup connection.

[dikkiedirk]

18 minutes ago, trurl said:

Did you really mean 900GB? 900MB in 8 hours sounds like a bad dialup connection.

900 GB it is! Sorry. Now at 1.24 TB of 2 TB. Disk is 99.9 % full.

 

Edited March 2, 2017 by dikkiedirk

[dikkiedirk]

Syslog looks clean. Rsync shows speeds between 30MB/s and 65 MB/s. It are older generation disks like WD20EARX.

[trurl]

7 minutes ago, dikkiedirk said:

Syslog looks clean. Rsync shows speeds between 30MB/s and 65 MB/s. It are older generation disks like WD20EARX.

If you are reading a drive in the parity array and writing another drive in the parity array then I would say these are pretty good results.

[RobJ]

I agree.  Those speeds are reasonable.

[dikkiedirk]

At point 15 of the guide I am asked to change ReiserFS to XFS and vice versa. But what if the format box is auto? It may be repeated to double check the parity valid box is checked.

 

The message in red at the top appears in a different place than the checkbox

If you have more than 10 drives in the array or a lower res screen you might not see it immediately.

 

Can this also be done without without setting a new config and swapping disks? Cant you just stop the array, change format of the new empty disk to xfs, start the array, let the empty disk be formatted and start a new copy cycle? Or will parity always be broken this way?

Edited March 3, 2017 by dikkiedirk

[trurl]

4 hours ago, dikkiedirk said:

At point 15 of the guide I am asked to change ReiserFS to XFS and vice versa. But what if the format box is auto? It may be repeated to double check the parity valid box is checked.

 

The message in red at the top appears in a different place than the checkbox

If you have more than 10 drives in the array or a lower res screen you might not see it immediately.

 

Can this also be done without without setting a new config and swapping disks? Cant you just stop the array, change format of the new empty disk to xfs, start the array, let the empty disk be formatted and start a new copy cycle? Or will parity always be broken this way?

If you don't need to swap disks you don't need to new config. And even if you do swap disks, step 16 tells you to check the parity valid box, so even then, parity isn't rebuilt and isn't broken, except as noted at the top of the procedure regarding parity2.