06-23-2011 02:20 AM
Hi
I have some issues with a new Linux media-server.
Backup is fine, but it takes 5 minutes or more berfore the actual backup starts. If I backup through the master server, the same backup starts immediatly.
Netbackup version 7.1
Logfile from linux-server
....
11:00:46.204 [5134] <2> logconnections: BPDBM CONNECT FROM 197.17.247.237.37801 TO 197.17.247.37.1556 fd = 6
11:00:46.267 [5134] <2> db_end: Need to collect reply
11:00:46.484 [5134] <2> bpbrm wait_for_child: start <------------ Here it waits long time
11:06:50.426 [5134] <2> bpbrm mm_sig: received ready signal from media manager <------ Even here it waits !
11:09:51.127 [5134] <2> bpbrm wait_for_child: child exit_status = 0 signal_status = 0
11:09:51.127 [5134] <2> bpbrm wait_for_child: child exit normal
....
Can anyone help me why it waits
Regadrs
Lasse Forslund
06-23-2011 03:21 AM
Do you have name resoloution (DNS) in place ?.
06-23-2011 03:29 AM
Yes
Both lookup and revese-lookup works
//Lasse
06-23-2011 04:18 AM
Seems to me that the delay is between bpbrm and bptm on the media server.
Please enable bpbrm and bptm logs on the media server and see if you find anything useful in them after the next attempt.
11:00:46.484 [5134] <2> bpbrm wait_for_child: start <------------ child here is bptm
11:06:50.426 [5134] <2> bpbrm mm_sig: received ready signal from media manager <------ bptm...
06-23-2011 04:23 AM
As it is a new media server it will try checking all the interfaces to connect which will take time.
Try giving Required Interface on Media server Host Properties.
--Praveen
06-23-2011 04:38 AM
The storage on media-server is NFS mounted (Data_domain). We tested a backup to local disk on media-server. No delays.
Master server (Windows) has the same storage, CIFS-mounted, but without delays.
I will check the logs.
Thanks
Lasse
06-23-2011 05:25 AM
Probably an O/S related problem then since Windows is quick and Linux isn't. Who would have thought. Check with EMC if there are tuneables for Linux accessing the DD. Also check if you get the same speeds when the actual backup starts to write (from both master and media).
06-23-2011 07:25 AM
I have an O-L-D DataDomain doc with the following under the Linux heading. Not sure if it is still valid...
Linux Backup Servers
A Linux NetBackup Media or Master Server needs to mount restorer directories using NFS. On the backup server:
• Create a mount point (directory) such as /dd/rstr01/backup and create an administrative mount point, such as /dd/rstr01/ddvar
• NFS mount the directories on the new mount points:
mount -T nfs -o hard,intr,nfsvers=3,tcp,rsize=32768,wsize=32768,bg rstr01:/backup /dd/rstr01/backup
mount -T nfs -o hard,intr,nfsvers=3,tcp,rsize=32768,wsize=32768,bg rstr01:/ddvar /dd/rstr01/ddvar
• Add the following lines to the file /etc/fstab. The lines mount the directories at every reboot.
restorer:/backup /dd/rstr01/backup nfs hard,intr,vers=3,tcp,rsize=32768,wsize=32768,bg 1 1
restorer:/ddvar /dd/rstr01/ddvar nfs hard,intr,vers=3,tcp,rsize=32768,wsize=32768,bg 1 1
• For systems with at least 2 GB of physical memory, add the following lines to the file
/etc/rc.d/rc.local. The lines increase the amount of shared memory available to NetBackup.
echo 536870912 > /proc/sys/kernel/shmmax
echo 536870912 > /proc/sys/kernel/shmall
07-25-2011 11:47 PM
Hi again.
Back from vaccation :-).
I already have the setting you suggested. I have also changed NUMBER_DATA_BUFFER to 96. Now I got a lot of IOwait against my NFS-share.
Should I change SIZE_DATA_BUFFERS to reflect "rsize=32768,wsize=32768" (my mount options) It is set to 262144
Regards
Lasse