Recently we have had several users experience data loss and/or iFolder
data relocation to the recycle bin and to the iFolder conflict bin without
user interaction.

iFolder in our environment is clustered but is not shutting down
successfully 35% of the time when failed over or possibly due to server
abends and is auto-checking the user accounts on reload:

For example:
5/25/07 14:25:56 Starting iFolder server, version 2.1.7 [05/12/06]
05/25/07 14:25:56 Configured LDAP was found ready to use.
05/25/07 14:25:58 iFolder Application Root: DATA:\apache2\iFolder\Server
05/25/07 14:25:58 iFolder Host OS: NetWare 6.50 Service Pack 6
05/25/07 14:25:58 iFolder Host WebServer: Apache/2.0.54 (NETWARE)
05/25/07 14:25:58 Last session did not shutdown properly!
05/25/07 14:25:58 Checking user accounts...

However these incomplete shutdowns do not preceed our instances when users
are experiencing iFolder data consistancy issues

The iFolder error log contains entries like those below for multiple users.

Thousands of instances of:
05/25/07 06:12:54 rfincher/home - DC 84b1 parent index fixed
05/25/07 06:12:54 rfincher/home - 7520.MAP deleted
05/25/07 06:12:54 rfincher/home - 7521.MAP deleted
05/25/07 06:12:54 rfincher/home - DC 7522 deleted!
05/25/07 06:12:54 rfincher/home - DC 7523 deleted!
05/25/07 06:13:00 rfincher/home - file 82e7 discarded
05/25/07 06:13:00 rfincher/home - new databaseId: 4656b6ac
05/25/07 06:15:23 rfincher/home - check user complete
05/25/07 06:15:24 rfincher/home - DIRCON.DAT - did: 0 checksum error
05/25/07 06:43:34 rfincher/home - DIRCON.DAT - did: 0 checksum error
05/25/07 07:39:35 rfincher/home - DC 84b1 parent index fixed
05/25/07 07:39:35 rfincher/home - new databaseId: 4656caf7
05/25/07 07:41:16 rfincher/home - check user complete

Most users involved have large iFolder space quotas but they are not all
burdening the service with heavy sync loads when the issues occur. Rather
the instances when the users experience the iFolder data consistancy and
loss issues seem to coinside with heavy and extended service activity with
a very small subset of identifiable users.

My suspicion is that the iFolder data indexing facilities for the iFolder
data for these users is getting skewed and and then repaired, resulting in
the data consistancy and loss issues, but I don't know why. I cannot find
the level of detail on the iFolder backend mechanics involved to determine
a root cause and subsequently implement a solution.

Any information you could provide about the iFolder .DAT files and how
they and the databaseid's are used would be most helpful.

I look forward to your response.

-- Lawrence

Lawrence Kearney
Novell NetWare 6 Master C.N.E.
Novell GroupWise 6 Specialist
Novell Nterprise C.L.E.
Cisco Certified Network Associate
CompTIA Project+ Associate