11-06-2012 01:17 AM
we use Data Protector 6.21. During incremental backup different boxes under Windows XP, 2003R2, 2008R2 under VMware ESX have errors with infinite directories according to DP logs. But there are no infinite looping directories nor file system errors.
Our configuration is VSS, Change Log Provider, Enhanced Incremental Backup. I think, CLP and/or Enhanced Incremental Backup is the problem, because DP can't use VSS under XP.
An example from a client.
The error messages on the cell manager for each subdir or file in \Programme\Sophos\AutoUpdate\Cache\savxp are:
[Warning] From: VBDA@XXX "C:" Time: 06.11.2012 02:11:46
Cannot open: ( Falscher Parameter. ) => not backed up.
The only error message on the client is:
06.11.2012 02:09:06 VBDA.1292.5152 ["/lib/ipc/ipc_ob2.c $Rev: 9228 $ $Date:: 2010-05-31 18:28:24":183] A.06.20 b400
[IpcGetPortRange] Error parsing "5556-5565"
Name of Object more than 1023 chars \Programme\Sophos\AutoUpdate\Cache\savxp\ClassFilt
Different directories are affected, user dirs and program dirs.
Could someone please help what's going on here?
Solved! Go to Solution.
11-06-2012 11:56 PM
DP produced a huge 4GB debug.log on a client with thousands of lines containing:
7.11.2012 00:27:52 VBDA.3764.3404 ["/da/dacmn/bdarec.c $Rev: 23898 $ $Date:: 2011-09-07 12:20:26":1030] A.06.20 b400
[?????:1033][RmWriteObjErr] DA_REC_OBJ_ERR has been generated for [\Dokumente und Einstellungen\XXX\Lokale Einstellungen\Anwendungsdaten\Microsoft\Internet Explorer\Custom Settings\Custom0\Custom0\Custom0\Custom0\Custom0\C
I can solve the problem by simple renaming Custom0 and rename it back to it's previous name. But during next incremental backup, other dirs are affected. It's not a permanent solution.
11-07-2012 08:50 AM - edited 11-07-2012 08:52 AM
It's very strange.
I think that you should log a call to HP Support.
Let's start with the classical questions ...
Are you using the latest patches ? (post the output of omnicheck -patches)
Could you try without the change log provider to see if it change something ?
Can you post the backup specification or the screenshots of the options specified in the specification ?
11-08-2012 04:16 AM
we don't use the latest patches. It's my next job to update DP. A colleague of mine has already downloaded the patches some weeks ago but hasn't installed them so far.
Patch level Patch description
DPWIN_00582 Core Component
DPWIN_00582 Core of Integrations component
DPWIN_00584 Cell Manager Component
DPWIN_00577 Disk Agent
DPWIN_00578 Media Agent
DPWIN_00576 User Interface
DPWIN_00573 Oracle Integration
DPWIN_00585 English Documentation (Guides, Help)
Number of patches found: 8.
The interesting part of our configuration with CLP enabled is:
DYNAMIC 1 5
} -protect weeks 4 -keepcatalog same_as_dataprotection
DEVICE "HP:Ultrium 4-SCSI_1_XXX"
WINFS "E: Teil1" server1.abc.de:"/E"
I disabled CLP on all clients during the last run of the job. No errors. It seems, that the CLP passes wrong names to the disk agent.
WINFS "E: Teil1" server1.abc.de:"/E"
I use DP Manager to configure the jobs. Usually I don't change the job files directly.
Interesting effect: when a disk agent's debug.log reaches 4 GB, the whole backup session with all running clients stops. You can delete this huge debug.log to continue the session. A client on a small XP-box with the Custom0-problem stopped the whole session at 4.5 m errors according to the error counter of the DP Manager.
11-08-2012 05:33 AM
So the next step is ... do you really need to use Change Log Provider ?
We had some speed issues. They are mostly solved. Nobody reads release notes. Our file server (NTFS, many, mostly small, files) is the first Disk Agent which is started in the job, and the last one, which finishes. The initial tree walk takes 1 hour, an incremental backup takes usually 3-4 hours. The file server is backuped alone half of the time, when all other clients already have finished. To pass the data at a high rate to the tape, I decided to enable CLP, because the tree walk during the backup is not necessary and the data stream is not intercepted - I think so, but I'm not sure.
What I did to solve the speed problem
- OB2IPCNOSENDRECVLOOP/OB2SHMIPC - the main problem
- 3 partitions of directories on the NTFS file server = different backup objects, which could be backuped in parallel, according to DP Performance White Paper
I think we can live without CLP - today. But I don't like to have this well promoted option available without the chance to use it.
11-08-2012 06:08 AM - edited 11-08-2012 06:09 AM
You should consider to upgrade to DP 7.01.
There is a lot of issue resolved with this release (ex: speed issue ;-)).
If your consider your issue as solved, mark the thread as solved. It may help some people that are looking for the same issue.
12-08-2012 10:48 AM
To finish this thread and report the last problem: the IDB becomes very large. We log directories and filenames during backup. When I export the IDB with omnidbutil -writedb, every unproblematic client's fnames file is 10-50MB. Clients with my problem have a size of 15-30GB. The IDB stores every directory name like \Custom0, \Custom0\Custom0, \Custom0\Custom0\Custom0 and so on, until the disk agent stops this nightmare at 1024 characters per directory. This happens for every file and subdirectory in this directory.
Your only chance to reduce the size of the IDB is to identify and unprotect the sessions with this problem, followed by purging fnames from IDB.