General Data ONTAP 6.5/6.5.1 reversion issues

General guidelines for reverting from Data ONTAP 6.5.1  

The following guidelines apply when reverting from Data ONTAP 6.5.1 to an earlier version:

• If you are reverting to a release prior to 6.5, you must disable any 6.5 features before reverting.
• You cannot revert if an upgrade is in progress.

Stale file handles on vFilers  

Data ONTAP 6.4 included a change in the Network File System (NFS) file handle format that is not backward compatible. If you have defined vFiler^TM logical units on your system and you upgrade to 6.4 or later and then revert to an earlier version, NFS clients might report stale file handles on the vFilers. To prevent this problem, unmount any NFS clients that are mounted on vFilers. You can remount them after reverting.

Reversion with larger inodes  

Starting with 6.5, when you upgrade, your gFiler is upgraded to support larger inodes. There are two instances when larger inodes can cause a reversion to an earlier release to fail:

• While the inode upgrade is still in progress
• When there is not enough disk space or free inodes in a volume

  Note
  Reverting to an earlier release can take several hours, depending on the number of total inodes and inodes used on your system.

To determine whether an inode upgrade is still in progress, complete the following steps.

priv set advanced

wafl scan status

Volume vol1: 

Scan id       Type of scan          progress                         

      2       inofile upgrade      849387 of 33554472

When more disk space or free inodes are needed for a reversion, you see a message similar to the following:

Volume vol1: There are not enough free inodes to revert the ACLs 
with nlinks over 65535.

A minimum of ### inodes needed (maximum ###).

Minimum and maximum inodes needed depends on the number of nlinks over 65,355.

You can free up inodes in a volume by deleting files or by moving files to a volume that has free inodes.

Excessive number of subdirectories and hard links must be moved or deleted  

If you have more than 65,533 subdirectories or 65,535 hard links, you must move the extra subdirectories to another directory and delete the extra hard links. When too many hard links exist during a reversion, you see an error message similar to the following:

Scanning all inodes in volume "vol1" for non-ACL inodes with nlinks 
over 65535 0% done 

inode 102, file has too high of an nlink (81519) 

inode 101, file has too high of an nlink (90001) 

0% done. 

Large directories can be reverted but cannot be made larger  

Starting with Data ONTAP 6.5, larger directories are allowed than in pre-6-5 versions. If you have a directory larger than 65,535 KB, when you revert to a pre-6.5 version of Data ONTAP, the directory is accessible and remains at its current size, but you cannot make it any larger.

Reverting entries in the /etc/exports file  

Starting with 6.5, when you upgrade to a new version of Data ONTAP, new access rules are applied to the /etc/exports file. If you revert to a pre-6.5 version of Data ONTAP, the /etc/exports file is scanned to determine whether it conforms to the pre-6.5 format. If not, the reversion does not proceed until the file is fixed.

To fix the file, you can edit it manually or use the exportfs -d command to revert the file. Using exportfs -d fixes most of the reversion issues. Some changes to the /etc/exports file are not fixed with the exportfs -d command, and cause a message to be sent to the console during a reversion. If any of these issues are not fixed, that line in the /etc/exports file does not load and access is denied for that resource. These include the following:

• Lines greater than 4,096 characters, excluding comments
• Entries that exceed 256 characters in either the -rw or -root options
• Too many entries in the -access option

Data ONTAP has no way to determine whether entries in an access list represent a netgroup or a host name; therefore, no checking is done on names during reversion. If netgroups exist in the -rw and -root options, you must manually edit the file to conform to the pre-6.5 format; otherwise, the export line containing the netgroup does not load and access is denied for that resource.

For more information about the /etc/exports file, see the na_exports.5 man page.

RAID-related errors  

Starting with Data ONTAP 6.5, updates made to RAID can sometimes cause problems when you revert to an earlier version, and the reversion might fail. The following is a list of possible problems and the resulting error messages that are sent to the console.

RAID label assimilation in 
progress. Wait for assimilation 
to complete before reverting.

Please wait one moment while 
RAID updates the Failed Disk 
Registry.

There is a sparecore disk 
present which cannot be 
reverted. Wait for the 
sparecore to complete before 
running revert.

Running a RAID label version that is too new  

When you install a pre-6.5 version of Data ONTAP, you might get a message that the RAID label version you are running is either invalid or too new. This can happen when the gFiler contains LUNs that were previously upgraded with Data ONTAP 6.5, and were not reverted using the revert_to command. In cases where a LUN was not properly reverted, the following actions occur:

• The LUN is not assimilated.
• The LUN is moved to the broken LUN pool with "bad label" given as the reason.
• An Autosupport notice is sent to NetApp and to you.

Copyright © 1994-2004, Network Appliance, Inc. All rights reserved.
Part No. 210-00662
File: revert2.htm
Last updated: 04/12/04 17:11:59