www.mass.dk

Symptom:
A 5 minutes delay may be expired when Netbackup 6.x or 7.x writes to a EMC Data Domain NFS share.

21/2012 11:59:14 - connecting
02/21/2012 11:59:14 - connected; connect time: 0:00:00
02/21/2012 11:59:15 - Info bptm (pid=28467) start
02/21/2012 11:59:15 - Info bptm (pid=28467) using 131072 data buffer size
02/21/2012 11:59:15 - Info bptm (pid=28467) using 128 data buffers
02/21/2012 11:59:15 - Info bptm (pid=28467) start backup
02/21/2012 12:05:16 - begin writing
02/21/2012 12:05:38 - Info bpbkar (pid=28466) bpbkar waited 4885 times for empty buffer, delayed 4917 times
02/21/2012 12:05:38 - Info bptm (pid=28467) waited for full buffer 1 times, delayed 28 times

Cause:
Incorrect NFS mount options. Netbackup debug logs (BPMT) do not give any indication of why this delay is present even with VERBOSE = 5.

Resolution:
EMC recommend the following qualifiers (Kernel 2.6 or newer):

mount -t nfs -o hard,intr,nfsvers=3,tcp,bg rstr01:/backup /dd/backup

or /etc/fstab options:

duplo01:/data/col1/backup2 /backup2   nfs     nolock,hard,intr,nfsvers=3,tcp,retry=10,rsize=32768,wsize=32768,bg 0 0

By bitter experience I have noticed that the installation of HP-UX patch PHKL_41700 causes a severe drop in backup performance (up to 2/3).

According to HP the following workaround can be used. The kernel settings enable high resolution timers support

# kctune  hires_timeout_enable = 1

Document link to HP documentation for hires_timeout_enable : http://h20000.www2.hp.com/bc/docs/support/SupportManual/c02255648/c02255648.pdf

Better option is to back out PHKL_41700 if possible.

Update:

The "kctune hires_timeout_enable" workaround seems to do it's job. The official fix is located in Patch PHKL_41967 (require reboot).

Symptom:
Creation of a MSDP pool using the Wizard fails on SUSE 11. A inspection of the pdde-config.log (located in {storage_path/log/pdde-config.log) contain a "su: incorrect password". A su from root to the pddb user is denied.