UNSOLVED
D
david_knapp
1 Rookie
•
122 Posts
0
3442
October 21st, 2011 07:00
Very unreliable SQL backups
Here is environment - backing up W2K SQL DBs on clustered W2K platform. Using Networker 7.6.1. The DBs are fairly small, the largest is about 40GB, Most are 3-10GB. Have tried parallelism settings from 1 to 4 on the client. One day the backups will work, another day they will all fail with I/O write errors on the Legato virtual device showing in the SQL server log. This will happen right as the job is started manually, before any tapes get loaded. The job will seem to just hang in the Monitor tab showing Starting session:state queued, but will eventually show as failed. Yesterday they worked fine running in the default group, today they all fail more or less instantly. There is only one instance of SQL on this cluster. Sometimes they work fine, other times they just fail.
The SQL log shows the following messages:
Internal I/O request 0x2BEA1C58: Op: Write, pBuffer: 0x12E10200, Size: 512, Position: 0, UMS: Internal: 0x0, InternalHigh: 0x0, Offset: 0x0, OffsetHigh: 0x0, m_buf: 0x00B54154, m_len: 1, m_actualBytes: 0, m_errcode: 995, BackupFile: Legato#9957e911-b306-4dec-9515-5c1c2de3ed1d
BACKUP failed to complete the command BACKUP database [TestDB] TO virtual_device='Legato#9957e911-b306-4dec-9515-5c1c2de3ed1d' WITH name=N'LegatoNWMSQL'
BackupMedium::ReportIoError: write failure on backup device 'Legato#9957e911-b306-4dec-9515-5c1c2de3ed1d'. Operating system error 995(The I/O operation has been aborted because of either a thread exit or an application request.).
BackupVirtualDeviceFile::RequestDurableMedia: Flush failure on backup device 'Legato#9957e911-b306-4dec-9515-5c1c2de3ed1d'. Operating system error 995(The I/O operation has been aborted because of either a thread exit or an application request.).
The GUI is v5.2.1.299 on the client side. Sometimes the GUI can log in fine and run backups, other times it cannot start up and gives the following messages - SQL Server Provider Error, DBNETLIB[ConnectionOpen (Connect())]. SQL Server does not exist or access denied. Cannot log into the SQL Server Default Instance.
The command in the client is - nsrsqlsv -a virtualname-sql -R -v. Where 'virtualname' is the name of our cluster. I can ping the cluster and the backup server from both sides.
I think this is some sort of authentication problem, but it has me stumped.
david_knapp
1 Rookie
•
122 Posts
317
0
Posted October 24th, 2011 06:00
Also, this in the monitor log:
52775 1319462562 5 0 0 3984 3948 0 servername.us.xxx.com (pid3948) 23 COM Error. Code = %08lx 1 1 11 -2147217887
david_knapp
1 Rookie
•
122 Posts
317
0
Posted October 24th, 2011 11:00
I am the SQL admin as well. I tried running from the client, same errors. This was working for a while, then started getting these errors. I ran several full backups from the default group last week, then moved client to another group. They started failing then, although that may have been a coincidence. Runs from the default group no longer work either. When running from the server, it does not recognize the cluster name, even though the command is "nsrsqlsv - a virtualname-sql -R -v". The virtualname is correct. We get
savegrp: suppressed 1 lines of output.
29100:(pid 1376): Invalid SQL virtual server name virtuaname-sql
38563:(pid 1376): Invalid option -aDo I need to create a client with the virtual name as well as the individual node names? I tried that but was not successful.
We have Power Edition. That is supposed to have two cluster connections with it, but it does not look like they are being used, although I cannot really tell. The license conformance summary does not mention them.
Well, as SQL admin you surprised me with -R as MS itself does not recommend that (it kills PIT restore). -v you can also skip. I believe you use GA module version. Try to get from support build 358. I use that one and had never problems with it (not that I imply there is a problem, but... you get an idea). In my case, almost all SQL machines are VMs, but I have one physical cluster. What I use as backup command is:
nsrsqlsv -c -a SQLCLUSTERNAME
For -c I use FQDN as I use always FQDN for client names in NW. -a is shortname of course representing SQL cluster name.
I do not use any special user for this backup as local system is authorized as SQL admin.
david_knapp
1 Rookie
•
122 Posts
317
0
Posted October 21st, 2011 09:00
I get this from the daemon log on the client when I try to use the GUI - nsrexecd 215 Authentication Warning: The user's OS groups cannot be verified and will not be included in the credential. This means that remote hosts may not see the user as being a member of any OS groups. Domain: %s, User: %s. 2 0 7 MYDOMAIN 13 7 d.knapp.