RTO issue

[Vishal_1980]

Dear expert,

We are  facing RTO issues on two of our SAP Application severs App1 and App2  during Peak Load hours ( Month end in particular ).

We have done the port reset activity on App1 by removing the network cable from one port at a time and we noticed that when the app1 was operational with only port it was running fine without any RTO and simultaneously the RTO Issue was happening on the other App2 server where it was working with both LAN port ( Teaming Enabled ).  

The HP support suggest us :

"understood  that your network team has asked to change the APA configuration from Active –Active to Active –Passive .

At present APA is  configured in MANUAL  mode and you can change it to LAN_MONITOR (Active –Passive)"

Please suggest me on this.If any once can share me the steps will be an added advantage.If any inputs require please suggest.

Thanks & regards,

Vishal

P.S. This thread has been moved from -UX>System Administration to HP-UX > networking. -HP Forum Moderator

[Patrick Wallek]

>>We are  facing RTO issues

Perhaps it is a lack of caffeine this morning, but what are "RTO issues?"  What exactly are you talking about?

[Vishal_1980]

Hello Patrik,

I am talking about retransmission timeouts.

The issue is we are facing the ping drop on our 2 quality servers.It only happns at the peak hours.Those servers are access by many users in the production.We check all the primary stuff i.e network port ,cable.Even the connectivity.Every thing is working fine.

Therefore it hapns at only peak hour.

Please suggest me on this..If you require any stats please do suggest.

Thanks & have a great day ahead.

Regards,
Vishal

[Bill Hassell]

ping is a UDP message which sits at the bottom of TCP/IP protocol priority. So a single ping that is dropped is not unusual at all on a busy network. ping is certainly not a reliable measure of network load or even the functionality of a remote system. Multiple pings where the return time is widely varying would indicate some network congestion or possible router delays but this is something that requires interpretation and much more sophisticated tools to determine if there should be any concern.

Now if you're measuring actual RTO values, are they specific to a single server? Be sure to run lanadmin -g # on all your HP-UX servers (where # is the LAN number as in lan0 or lan3). The last half of the statistics (after Index) must be zero. If there are non-zero values, make sure you don't have hardcoded duplex settings (autonegotiate is mandatory for Gbit links, strongly recommended for 100 Mbit). On any ports where the error stats are non-zero, run lanadmin -x # to see the duplex and negotiation. Anything at half-duplex is wrong (except 10 Mbits which is virtually obsolete) and needs switch and server changes.

[Akio Kabutogi]

Do you see the counter shown below increasing ?

# netstat -s |grep 'retransmit timeout'
        612 retransmit timeouts

As Bill said, ping relies on the protocol (ICMP) below TCP. 

Some packet lost at ping level may be an indication for TCP's retransmits.

But I would try to monitor the counter shown above to be sure.

Also,  the counter shown below is helpful.

# netstat -s |grep 'duplicate acks'
                75244 duplicate acks

If this is also increasing, that implies "We sent a packet but haven't received ack to it", which indirectly

show there are some packets dropping at link layer, the wire or the switch.

You may want to check the switch'es statistic couner as well.

[Vishal_1980]

Hello Akio,

Thanks for the revert i will check this whatever you said & get back to you very soon.

Thanks & regards,

Vishal

[Vishal_1980]

Hello Akio,

I tried the same that you ask.

Thing is that the proble occurs only at the end of the month where we generally have a load on the server

& on the network as well.

Please find the attached out put of the command & i will requiest you to please suggest em opn this.

Like is this issue is related to HP or completely related to Network.

THanks & have agreat day ahead....

Regards

Vishal