using ports of the same network interface
Hi,
I'm an Oracle DBA.
I will perform 2 installations of Oracle Management Agent on a Solaris server with the following config:
SunOS 5.10
Interface IP AddressHostname
- --
ce6address0hostname0
ce6:1address1hostname1
ce6:2address2hostname2
The first installation will be done on <hostname1>.
The second one will be done on <hostname2>.
Each agent will use a port that defaults to 3872.
Can I use the same port 3872 for both installations?
Thanks and regards,
Giuseppe Centioli
# 1
Yes, as long as the listener knows to bind only to the particular IP address in question.
This is what allows cluster failover solutions to work.
If it did not bind to the particular IP address and instead took the default path of binding to "any" address, then the first one to come up would grab the port on all address and the second one would fail.
--
Darren
# 2
Hi Darren.
The 2 installations will run at the same time in a sigle server.
Each one need a dedicated port.
I am concerned because hostname1 and hostname2 use the same phisical interface ce6.
ce6:1 and ce6:2 are logical interfaces of ce6.
I tested the following:
I configured 2 oracle listeners on this server.
The first one uses hostname1 and port 1831
The second one uses hostname2 and port 1831.
I started the first one successfuly.
When I tried to start the second one I got an error.
How does Solaris manage this scenario?
Regards,
Giuseppe
# 3
> Hi Darren.
> The 2 installations will run at the same time in a
> sigle server.
> Each one need a dedicated port.
> I am concerned because hostname1 and hostname2 use
> the same phisical interface ce6.
> ce6:1 and ce6:2 are logical interfaces of ce6.
That doesn't matter. The application can't tell the difference between addresses that share an interface and addresses that have their own.
> I tested the following:
> I configured 2 oracle listeners on this server.
> The first one uses hostname1 and port 1831
> The second one uses hostname2 and port 1831.
>
> I started the first one successfuly.
> When I tried to start the second one I got an error.
What error? Before you started the second one, what did
'netstat -an | grep 1831' show?
> How does Solaris manage this scenario?
Exactly like I mentioned before. As long as the binding does not overlap, it should be allowed.
--
Darren
# 4
Hi Darren.
netstat -an | grep 1831
<hostname1>.1831 <hostnameX>.1399 639800 504000 ESTABLISHED
*.1831*.*00 491520 LISTEN
The error is an Oracle related error telling that another LISTENER
has been started with port 1831.
Thanks,
Giuseppe
# 5
> Hi Darren.
>
> netstat -an | grep 1831
> <hostname1>.1831 <hostnameX>.1399 639800
> 504000 ESTABLISHED
> *.1831*.*00
> 491520 LISTEN
Okay, that's incorrect. You'll see that the process is listening on *.1831. In other words, it's listening on *all* network addresses. That was the specific point I was making earlier. It must not do that. The listener must bind only to the network address chosen.
When it's set up right, you should see a specific address on the LISTEN line, not a *.
> The error is an Oracle related error telling
> that another LISTENER
> has been started with port 1831.
Yes, that's the case here. The listener is not binding to the correct address in your configuration. I am not an Oracle DBA, and I can't tell you what the correct syntax is.
Hmm. Okay, here's some stuff from google. Do you have a hostname specified in your listener.ora file?
(ADDRESS_LIST =
(ADDRESS = (PROTOCOL = TCP)(HOST = hostname)(PORT = 1521))
Like that?
--
Darren
# 6
Darren,you're right. LISTENER1 is configured with HOST=<hostname>.The correct HOST for LISTENER1 is <hostname1> .It works now!Thanks a lot.Have a nice week-end.Giuseppe
