Link to home
Start Free TrialLog in
Avatar of omarfarid
omarfaridFlag for United Arab Emirates

asked on

Oracle 10g RAC database TCP parameters on solaris 10

Please provide optimized TCP parameters on Soraris 10 (SPARC) for OLTP Oracle 10g RAC database for below TCP parameters

tcp_rexmit_interval_min
tcp_rexmit_interval_max
tcp_rexmit_interval_initial
tcp_ip_abort_interval
Avatar of Brian Utterback
Brian Utterback
Flag of United States of America image

Generally you should leave these at the defaults. However, if you are on a closed LAN with very small round trip times they can be tuned to be more efficient. The latest kernel update patch for Solaris 10 already reduces the tcp_rexmit_interval_initial parameter from 3 seconds to 1 second. You could go down to 500ms, but I wouldn't go lower than that. On the same token, you could reduce the tcp_rexmit_interval_min to 200ms. You should never reduce the tcp_ip_abort_interval below 1 minute, but the abort interval doesn't affect the transaction efficiency anyway.
Avatar of omarfarid

ASKER

Thank you blu, can you provide more info on how to tune these parameters? Is there a document covering their tuning?

currently, my servers parameters values are as below:

h1 (sun v480):
#########
tcp_rexmit_interval_min=400
tcp_rexmit_interval_max=60000
tcp_rexmit_interval_initial=1000
tcp_ip_abort_interval=300000


h2 (sun v480):
#########
tcp_rexmit_interval_min=400
tcp_rexmit_interval_max=60000
tcp_rexmit_interval_initial=1000
tcp_ip_abort_interval=300000

z1 (sun v490):
########
tcp_rexmit_interval_min=400
tcp_rexmit_interval_max=60000
tcp_rexmit_interval_initial=3000
tcp_ip_abort_interval=480000

z2 (sun v480):
########
tcp_rexmit_interval_min=400
tcp_rexmit_interval_max=60000
tcp_rexmit_interval_initial=1000
tcp_ip_abort_interval=300000

Oracle support said "these parameter depends on customer environment and can vary up the h/w , network topologies, switch etc" but did not give any guidelines on how to set them or what are the recommended values.

This is related to radius restarts and we are suspecting connectivity issue with the oracle RAC database
ASKER CERTIFIED SOLUTION
Avatar of Brian Utterback
Brian Utterback
Flag of United States of America image

Link to home
membership
This solution is only available to members.
To access this solution, you must be a member of Experts Exchange.
Start Free Trial
As I said, we are facing radius server crashes and suspection is twards database. There is a firewall in between the database RAC servers and the RADIUS server.
What I mean is, these prameters are dealing with the retransmissions of dropped packets at various stages of the connection. Is the problem you are experiencing related to dropped packets and aborted connections? Will faster or slower retransmissions help? Shorter or longer abort timeout?
How to check if there are dropped packets on the servers themselves?
You can use netstat to see the retransmission counters for TCP.
Thanks for your support. The case is still open with Oracle