Performance issue while taking Archive from same TSM server and TSM client

[aixtsmsan]

Hi

I am taking 1TB of archive of filesystems which are there on TSM server itself. I am using gigabit IP address from earlier 100Mbps IP address inorder to perform faster archives by mentioning the same gigabit IP address in the dsm.sys options file.

Before: Only TCPServeraddress option was there in dsm.sys file
CommMethod: TCPIP
TCPServeraddress: 10.200.99.88 (100Mbps NIC card)

Now:
COmmMethod: TCPIP
TCPSErveraddress: 10.100.90.80 (1Gbps NIC Card)
TCPClientaddress: 10.100.90.80

TSM Server and TSM client is on the same server.

TSM Version: 5.5
Operating System: AIX 5.3

However after couple of weeks, backup/archive sessions only for that Client became slow. Even dsmadmc command is used to connect TSM server is taking long time to prompt for password and after loging, output for any commands entered are taking long time to display. However I do not see the probelm connecting TSM server using Web Interface and connecting to the TSM server from another Workstation using dsmadmc.

after rebooting AIX Server, issue does not exists and again appear after 10 days later.

do I need to set any N/W settings on NIC cards or any other options I need to change in dsm.sys file for taking backup/archive from the same TSM server and TSM client.

 

[Harry_Redl]

Hi,

strange - anything in the logs? As you said the client and the server are on the same machine - what about using COMMMETOD SHAREDMEM?

Harry

 

[aixtsmsan]

Thanks for reply Harry.

I did not find anything on the logs. However when I try to ping the IP addresses: The Time is showing as 1000ms in the ping request.

Before Reboot:

PING tsmserverip: (10.100.90.80): 56 data bytes
64 bytes from 10.100.90.80: icmp_seq=0 ttl=255 time=1002 ms
64 bytes from 10.100.90.80: icmp_seq=1 ttl=255 time=1008 ms
64 bytes from 10.100.90.80: icmp_seq=2 ttl=255 time=1005 ms
----tsmserverip PING Statistics----
4 packets transmitted, 3 packets received, 25% packet loss

After reboot:
PING tsmserverip: (10.100.90.80): 56 data bytes
64 bytes from 10.100.90.80: icmp_seq=0 ttl=255 time=0 ms
64 bytes from 10.100.90.80: icmp_seq=1 ttl=255 time=0 ms
64 bytes from 10.100.90.80: icmp_seq=2 ttl=255 time=0 ms
64 bytes from 10.100.90.80: icmp_seq=2 ttl=255 time=0 ms
----tsmserverip PING Statistics----
4 packets transmitted, 4 packets received, 0% packet loss


I had already rebooted this server more than 6 times from past 2 months to clear the problem. I had googled about sharedmem option but I am not sure whether It may resolve the problem. I do not want reboot once again. Hence I left the options to simple which was working earlier however it was slow.

Commmethod: TCPIP
TCPServeraddress: 10.200.99.88 ((100Mbps NIC card)
TCPClientaddress: (I had removed this option from dsm.sys)

 

[domino]

try TCPSErveraddress: localhost

 

[aixtsmsan]

Thanks Domino.

Honestly I cannot try any further until I get clear root cause of the problem. I had already rebooted the server number of times. I had raised a call with IBM. They suggested to upgrade IBM TSM client from 5.4 to 5.5.3. I upgraded it.

 

[Harry_Redl]

Hi,

I think it is an AIX problem, not TSM problem. You have high "ping" latencies from the OS - and this is outside of TSM.
Other people are having this problem
http://www.ibm.com/developerworks/forums/thread.jspa?messageID=14413315
Need an AIX expert here ...

Harry

 

[aixtsmsan]

I do not see any AIX issue. It may be N/W problem. However the IP addresses are used by application for data transfer. I had left options in dsm.sys file to simple.

I will follow up with IBM and N/W team in our company regarding this. I will update you.

 

[tariq.kaifi]

Hi,

I can understand that you do not want to reboot the server, but I guess domino wanted to figure out if there is any problem in the setting in your network card maybe. Since, it dont require any sort of reboot, you may try to use "localhost" or loop back ip "127.0.0.1". You can revert if it does not help.

By the way, we will wait to hear IBM's and network's say as well.

 

[aixtsmsan]

Hi Tariq,

I see latency on loop backup address as well. IBM has suggested to change n/w option on AIX no -a lo_perf=1(default) to 0 and reboot the server to reflect the new settings. However I did not get the root cause of the problem. IBM level2 has suggested if the problem occurs, they may have to collect snap and escalate to level3 support for further analsys.

#no -ro lo_perf=0

PING 127.0.0.1: (127.0.0.1): 56 data bytes
64 bytes from 127.0.0.1: icmp_seq=0 ttl=255 time=1000 ms
64 bytes from 127.0.0.1: icmp_seq=1 ttl=255 time=1002 ms
64 bytes from 127.0.0.1: icmp_seq=2 ttl=255 time=1001 ms
----127.0.0.1 PING Statistics----
4 packets transmitted, 3 packets received, 25% packet loss
round-trip min/avg/max = 1000/1001/1002 ms

 

[aixtsmsan]

I am still not able to figure out whether it is a AIX or TSM problem...I guess it could be AIX problem as Harry also updated same thing in the below comment.

Thanks Harry.