cancel
Showing results for 
Search instead for 
Did you mean: 

Slow data stream on RedHat 5.8 with Netbackup 7.5.0.3

Daniel_Jastredo
Level 4

All,

I have a problem with a slow data stream when running a Netbackup Media Server 7.5.0.3 on RedHat 5.8. We are migrating systems from Solaris 10 to RedHat 5.8. The speed on Solaris 10 is about twice as fast as the RedHat 5.8 system. Also, we have no problem with earlier versions of RedHat, just 5.8. I was told that the 5.8 version is a different animal than even RedHat 5.6. Is there a different OS config I need to set up to get RedHat 5.8 and Netbackup 7.5 to get along? Possibly buffer changes, kernel tweaks, etc?

Thanks.

 

7 REPLIES 7

Nicolai
Moderator
Moderator
Partner    VIP   

1: check SIZE_DATA_BUFFERS & NUMBER_DATA_BUFFERS in Netbackup

2: Check NIC is working as intended. Use the Netbackup GEN_DATA directive to verify path can transfer data. http://www.symantec.com/docs/TECH75213

Am using RH 5.8 and has no problem in trasfering >300MB/sec via 10GB ethernet

Daniel_Jastredo
Level 4

This backup is going over the SAN to LTO tape. Besides the buffers, any other tweaks?

Nicolai
Moderator
Moderator
Partner    VIP   

There are special buffer for FTL configurations (NUMBER_DATA_BUFFERS_FT & SIZE_DATA_BUFFERS_FT ) . Do you got those set ?

 
You may also have a marginal good tape drive. Marginal good tape drives does not reach the maximal writing speed because of write corrections. Try using a cleaning tape
 
If the client is UNIX type check out the GEN_DATA directive (See my previous post)  in combination with a /dev/null disk stu on the media server. If this does not rock I believe you got a SAN problem then.
 
http://www.symantec.com/docs/HOWTO56132

Daniel_Jastredo
Level 4

The "FT" buffers wouldn't apply in my case since I'm not using Fiber Transport, but I will double check the standard I/O buffer settings. What I'm looking for is info on the RedHat 5.8 O.S. vs Netbackup. Are there any tweaks I need to make with this config that wasn't needed in RH 5.6 or Soalris 10?

 

Thanks.

Nicolai
Moderator
Moderator
Partner    VIP   

None that I am aware off. The only teaks we do is some TCP/IP settings. The rest is left to default

Add the following values to /etc/sysctl.conf

kernel.msgmni = 256
kernel.sem = 300 32000 32 1024
# Make sure TCP sessions are kept alive
net.ipv4.tcp_keepalive_time=1800
# Recommended network settings.
net.core.rmem_max = 16777216
net.core.wmem_max = 16777216
net.core.rmem_default = 262144
net.core.wmem_default = 262144
net.ipv4.tcp_rmem = 8192 524288 16777216
net.ipv4.tcp_wmem = 8192 524288 16777216
net.core.netdev_max_backlog = 30000
 
Activate sysctl settings with command:  (A system reboot will automatic read and apply /etc/sysctl.conf )
sysctl –p

Daniel_Jastredo
Level 4

Sounds good, thanks.

AAlmroth
Level 6
Partner Accredited

On RHEL you may have other services that may impact the general performance, to name two: SELinux and iptables. If these security services are active, and you have very heavy network load, as in the case of a media server, you may see noticably lower performance.

Nicolai's network settings will definately improve the TCP performance. Another part to look into is the settings of the net driver/NIC HW. You can try with/without TCP checksum offloading etc.

/A