This article will discuss the DP client’s performance when execute a backup job.
OS:
Client: RedHat Enterprise Linux 5.4 x86
Server: Windows server 2003 R2 x64
Software: HP Data Protector A.06.20 b370
Storage: iSCSI Storage (poor performance disks)
Client Server: HP BL680 G5 with 16 cores,16GB RAM.
Symptom: When a backup job is issued, the load of the server climbed up from 0.xx to more than 10.xx, especially the utilization of CPU of 100% , which has a bad effect of the Oracle service.
Solutions:
Firstly, we take a look at the load when a backup is running.
Then , when I kill the process of “omni”, we can see this:
Then , when I kill the process of “inet”, we can see this:
So the DP has caused a big problem!
When I invested the log of DP’s agent, I found so many entries like this:
Which tells the server is not the Cell Manager of the client.
When I checked with the HP service, it is said that may DNS cause the problem.
Even my /etc/hosts has relevant entries.
The solution is quite simple as follow:
Just add an entry of Cell manager’s host name.
But in some systems there may be no such file, so I think we should create it manually.
Ok, here we’ll see the load of the backup job.
The “vbda” process is the backup process, but too odd to see “omni” and “inet” process!
HP told that we should add the cell manager hostname entry into the /usr/omni/client/config/ allow_hosts, but I didn’t find such directory and file. So I did it in the /etc/opt/omni/client,and touch a new file of “allow_hosts” .
The process view:
Always Backup Restore Media Management Cell Manager Windows omniinet.exe rds.exe mmd.exe crs.exe uiproxy.exe kms.exe bsm.exe rsm.exe msm.exe UNIX rds mmd crs uiproxyd kms bsm rsm msm Disk Agent Client Windows omniinet.exe vbda.exe vrda.exe UNIX vbda vrda Media Agent Client Windows omniinet.exe bma.exe rma.exe mma.exe UNIX bma rma mma
Till now, I made the backup job ease on the system.