Index troubleshooting guide

L. Weissman
(15 Oct, 1998)

There are a number of possible circumstances which indicate index problems. 'WISS' errors are covered in a separate Action, but the solutions are often the same. For most index related error messages, the following troubleshooting is suggested:

(See Background information below)

First try to confirm that you actually are having index related problems. This will either be error messages related to the indexes or messages from index related processes, such as, nsrindexd, etc. Other symptoms might include disappearing or missing file entries in the recover screen, lost save sets or similar mysterious save or recovery problems.

Be sure sure that all patches that affect indexes have been installed. See Legato's web site for complete information for all versions. This is mandatory before proceeding.

If you suspect the media index may be corrupt, you can run
    nsrim -X
to run a consistency check the media index. It is generally a good idea to run this first, so that any problems in the media index do not adversely affect the file index cleanup (if needed). There is not a utility for repairing the media index, so if the media index is the problem you will probably need to do something resembling a disaster recovery. The media index is recovered with
    recoverindex (ver. before 4.2) or
    mmrecov (ver. 4.2 and later)
        ; see appropriate man pages

In NetWare, use the "NETUTIL" utility to recover indexes.
In NT, run "mmrecov" from command line (similar to UNIX v. 4.2)
In all cases, see your manual or man pages for more detailed information on index recovery.

After you are sure the media index is O.K., run nsrck -F to see if this will fix the problem. (nsrck -F will not repair the media index, it only works on file indexes - that's why we ran nsrim -X first). Networker must be shut down first.

If this doesn't fix the problem, you can try to recover a previous index. Client file indexes can be recovered like any other ordinary file on the server using the recover command or Recover GUI on the server. They will usually be located in the directory /nsr/index on the server.

Another possibility is to rename or delete the existing indexes and start fresh. This can also be a good troubleshooting idea. If problems disappear when a fresh index is used, its a reasonable choice that there may be index corruption. If an occasional recovery is necessary from older tapes, they can be scanned back into the indexes.

If you must have a complete prior index, scanning it in may be a long process. This is a fairly extreme step, but may be necessary if an uncorrupted version of the index cannot (easily) be recovered.

If, after applying the patches, fixes or suggestions provided above, the problem has not been fixed, please gather all necessary troubleshooting information and escalate this incident.

Background

NetWorker keeps two types of indexes

1. a media index - one (only) for each server, which keeps track of what is on each tape (or other media) that has been used by the server.
2. a file index - typically one for each client, which keeps track of what files and versions of files have been backed up for each client.

The entries in these indexes (and the size of the indexes) are managed automatically through the Browse and Retention Policies for the client and the server (See Tech bulletin 001 for more information)
These files can become corrupted in various ways (for example, if the system crashes while the file is open).

NetWorker attempts to correct problems with these indexes automatically, by running checks and/or repairs from time to time. However, this is not always successful.

Problems are often indicated by a number of error messages, often of the form "Unable to save index for client <client-name>" or other messages indicating the same idea.