PDA

View Full Version : Planner missing results



jours
July 21st, 2010, 08:20 AM
Hi all,

I have amanda set to back up three servers to Amazon. Everything seems to be working for most of the DLEs, but a few of them give me this error:



FAIL planner cle-winalba-01 "C:/SQLBackup" 20100721111312 0 "[missing result for \"C:/SQLBackup\" in cle-winalba-01 response]"
FAIL planner cle-winpool-01 "E:/Ed_Pulldown" 20100721111312 0 "[missing result for \"E:/Ed_Pulldown\" in cle-winpool-01 response]"
FAIL planner cle-winpool-01 "E:/SQLData/MSSQL10.MSSQLSERVER/MSSQL/Backup" 20100721111312 0 "[missing result for \"E:/SQLData/MSSQL10.MSSQLSERVER/MSSQL/Backup\" in cle-winpool-01 response]"


When I hit Google with this error, most of the results looked like catastrophic configuration errors where the client wouldn't communicate at all. But in my case other DLEs on the same servers back up fine:




cle-winpool-01 E:/SQLData/MSSQL10.MSSQLSERVER/MSSQL/Backup zwc-normal
cle-winpool-01 E:/Ed_Pulldown zwc-compress
cle-winpool-01 E:/hparoot zwc-compress
cle-winpool-01 E:/wwwroot zwc-compress
cle-winpool-01 F:/hparoot zwc-compress

cle-winalba-01 C:/SQLBackup zwc-normal
cle-winalba-01 C:/Inetpub zwc-compress


Curiously, the zwc logs on the windows boxes don't mention the failing path at all. So if I search the log file on cle-winalba-01 for "SQLBackup" I get nothing at all...even with debug level set to 5. (For what it's worth, I do see it being logged during an amcheck.)

Amcheck runs with no errors at all:



-sh-3.2$ amcheck DailyBackS3
Amanda Tape Server Host Check
-----------------------------
Holding disk /backup01/holding/DailyBackS3: 98 GB disk space available, using 0 GB as requested
slot 9: volume 'DailyBackS3-09'
Will write to volume 'DailyBackS3-09' in slot 9.
NOTE: skipping tape-writable test
Server check took 0.692 seconds

Amanda Backup Client Hosts Check
--------------------------------
Client check: 3 hosts checked in 3.825 seconds. 0 problems found.

(brought to you by Amanda 3.1.1)




So...where should I check next? Any help would be appreciated. Thanks!

John

jours
July 22nd, 2010, 06:51 AM
FWIW, I modified one of the windows hosts to use a a client template last night and it worked fine. So the client template on cle-winalba-01 has:

c:\Inetpub
c:\SQLBackup

and it runs with no errors. But when I try to do a recover, I see both those directories inside a single DLE...which means a restore needs to pull them both from the tape to restore one directory, right?

So would the best practice here be to put each directory in its own client template? Or is this even a viable workaround?

I'm inclined to think the initial issue of "missing result" is a bug of some sort since I can't find a failure in any of the logs...windows client, /etc/amanda, or /var/log/amanda.

John

paddy
July 24th, 2010, 07:29 AM
Does this mean you were able to perform successful backup with the template?

Restore will take more time because folders in the template are part of the same
template. Best practice is to put folders directly in the disklist file on the server. Use
templates only when you need consistent snapshot across folders or you need to
use pre-/post- backup or restore operations.

"missing result" might be a bug. Are you running Windows client 3.1.1?
Paddy

jours
July 29th, 2010, 10:39 AM
Yes, I am able to backup successfully with the templates. But as you said, I don't really need/want to use them if I don't have to because of the restore time...and it seems to play hell with incrementals as well.

I am running ZMC-Community-3.1.1-32bit on the troublesome Windows servers. Thanks!

jours
July 30th, 2010, 10:34 AM
Update... I let the backups run twice with the windows client templates and they work fine. That's not going to be workable because I can't get incrementals that way.

So I went back to putting each directory in the disklist explicitly. The backup still fails on that machine, but now it only fails on one directory E:/wwwroot, which wasn't failing before. The ones that were failing are now working!

Safe to assume this is a bug in the windows client? Or anyone have any other ideas?