Unix Solaris configuration with Oracle database

[marciob]

Environment information:
* Oracle version: Oracle 9i
* Operating system: Unix Solaris 9
Personal I need a lot of help from you I have an Oracle server with the settings below and I wonder if I changed the settings like Oracle and your connections will behave.

follows the settings I have today:

Select all

ndd -get /dev/tcp tcp_time_wait_interval 
60000 
 
 ndd -get /dev/tcp tcp_fin_wait_2_flush_interval 
675000 
 
 ndd -get /dev/tcp tcp_conn_req_max_q 
128 
 
ndd -get /dev/tcp tcp_keepalive_interval 
7200000 

follows information that advise:

Select all

ndd -get /dev/tcp tcp_time_wait_interval 
60000 
 
 ndd -get /dev/tcp tcp_fin_wait_2_flush_interval 
67500 
 
 ndd -get /dev/tcp tcp_conn_req_max_q 
8000 
 
ndd -get /dev/tcp tcp_keepalive_interval 
15000

I am waiting but I am worried about Oracle, so that I am taking That doubt with you.


Hugs

[pauloaleo]

Is your application going to be webserver? Java? Or simply a normal bank with client / server application?

[marciob]

Not the server will only be bank server and with ftpserver.


Only this

[pauloaleo]

Why the settings you put would be ideal for a WebServer, or IAS, etc.

Who indicated these settings for pure bank?

[marciob]

Well paulo my problem is the following

I have a server today with Oracle 9i and Solaris 9 on this server runs the FTP Server too, but something happened in my network, well I think that today There are several FTP connections being lost all randomly, and also several connections established from DNS. With this I searched the internet and on a website said "you are having problems with TCP connections in Solaris" and down ordered to carry out the procedures below. But as I have oracle connections on this machine, which are many tb, this wondering if this would disrupt the connections with Oracle if you perform these procedures.


I think it was clear.

[pauloaleo]

Just a question. Something happened for this problems appear or began "out of nowhere"?

[pauloaleo]

I am asked to try to establish what may be happening.

is Keep_Alive that makes FTP connections fall after a while, with errors similar to the following:

"Disconnected from server
Terminating Connection. "

or

"Disconnected From Server
Timeout (900 Seconds): Closing Control Connection."

Sometimes 2 hours is little time. Increasing this does not harm Oracle at all.

[marciob]

dude well the change of network has not yet been discovered because we can not know because who cared for this is another team .... but the FTP connections began to fall already has a couple of weeks and this happens randomly because time connects and Time does not connect, and the errors according to the log's log's are at different times, at the time of download before downloading the file, when it will list the file to make sure the file has been copied and etc.



What me and my team we are thinking is that it can be due to the large number of connections to the DNS that is making this happen, but still We do not continue to say or better know what is causing this large number of DNS connections. Just for you to have a idea we have today 1890 DNS connections on this server



about increasing this time of keep_alive, if I do not mistake you had asked me to accomplish this increase, with this, I decided to search on the subject and found a forum that the guy asks to increase all those parameters and then I decided to take this doubt, but Miha concern is with Oracle

[pauloaleo]

Well, I do not know if it will help, but take a look at the link:
http://www.sun.com/blueprints/0603/816-5240.pdf
I'vê seen such problems in AIX, but in the end, an ARP -D solved.

Have you seen if Solaris is solving the reverse of DNS?

[marciob]

no, no, no, no, no, no, no, no, no.


Face I will give you a read and return you. OK?


thanks for the help that you are giving me there

[pauloaleo]

See also this link
http://www.sun.com/bigadmin/content/sub ... g_tip.html
It has several DNS configuration demos in Solaris.

at least to check with existing DNS there.

[marciob]

Paulo from what I understand the sent link is so I can create and set up the DNS in Solaris .... but in fact my server is not a DNS nor can it be ....



I would like more information on how I can resolve with the "ARP -D"

would be able to guide me?

[pauloaleo]

The entire server has an ARP table. When you type

Select all

arp -d hostname 

it cleans this table. The ARP is Address Resolution Protocol and dynamically stores all IPS, MAC address, TTL, etc.

works like this:

Source node Query your ARP table
If the destination node is not in the table
Sends an ARP Query in diffusion
Destination node responds directly to the source node with an ARP Reply
Source node enters with the record in the table
if the destination node is in the table
Extracts the MAC address

follows the

Select all

$ arp 
Usage: arp hostname 
       arp -a 
       arp -d hostname (limpa a tabela) 
       arp -s hostname ether_addr [temp] [pub] [trail] 
       arp -f filename

All equipment on your network has an ARP table.

[marciob]

Paulo Good morning


Well just to leave you positioned, thanks to you I realized that I really have my Solaris servers are not cadated in DNS and so not doing the reverse .. . I already requested the registration in DNS by those responsible ...

well what I am doubt is when I type ARP -D "Hostname", well what is the hostname? My bank server or my DNS server that has several requests?

[pauloaleo]

All servers have ARP table. Even the switch has one.

I'm talking about the experience I'vê ever had. Once, in a company I worked, they made a maintenance in the Core switch, and all AIX machines of database stopped responding.

It took me a 3 hour, analyzing the problem and not getting anywhere. Until no one had warned me of maintenance. When you warned me, I went on each bank server is then the XXXXXXX ARP where the XXXXXXX was the bank server hostname.

Then everything started working normally.

[marciob]

Paul I performed the ARP -D procedure but when I perform the netstat command, several connections to DNS appears.
Some are established and other time_wait


would be to clean this table or to end the problem of the fall or the non-response of my TCP Connections?

[marciob]

Dude I can not access my bank machine after the command


What do I do?

[pauloaleo]

When I did in AIX, the table cleaning solved the connection problem.

You can no longer connect to the server?

[marciob]

Hi paulo,

I had to restart the server to function



now normalized but could not Telnet, FTP, nor access to the bank.


is now normal ....


but tb did not clean the netstat

99]]
Continues showing all connections.


Even so, thank you very much for the help you gave me several guidelines and one was very important that of the reverse that apparently it seems that solved the problem of my connections to fall


Thank you

[marciob]

Well folks I'll come back here again to say that nothing I got with my problem


The number of DNS requisitions on my bank server is high and is hurting my FTP

My bank servers have already been registered in the main DNS and are already performing the reverse tb

but the loses of FTP connections during copy of files It is very high


I do not know what else to do


I ask if anyone can help me


I need a lot

[victorhugomuniz]

You adjusted the time out ne ..

would not have any variable on the server to adjust the file size?

[pauloaleo]

The only thing that is left is Keep_Alive.

Have you raised this TCP parameter?