ROOTVG

Hi All,

Hope you can answer my question "How to know how many live TCP connections the AIX server will permit" Is this something I can view in the "no" command? thanks a lot in advance!

Take a look at the first section of
netstat -an
- for TCP sessions ESTABLISHED, etc.

If your reason for asking was that you are hitting a limit,
depending on what the TCP using application is, perhaps that's the PTY settings:
smit chgpty

Thanks a lot John,

The reason I ask is there are some various testing being done right now on the server. From the network's view, they mentioned that there are a lot of tcp server refuses connections. Do you think the PTY setting's is what causing the tcp refuses? Or could it be that the server is busy and it refuses more connection.

Thanks again.

The PTY settings are just one example limitation at the application level,
e.g. affecting concurrent telnet connections.

The server may refuse TCP connections for lower level reasons - e.g. not enough queues and buffers
in the tuning of the adapter, interface and kernel.  There's quite a bit to it.

The netstat statistics and/or entstat outputs might reveal queue/buffer config problems.

By the way, a good way to capture all the stats in a consistent manner,
around a problem like this, is to run a "perfpmr" - downloaded from here:

ftp://ftp.software.ibm.com/aix/tools/perftools/perfpmr/ (http://ftp://ftp.software.ibm.com/aix/tools/perftools/perfpmr/)

If you are having performance problems of any sort, this is something
we can analyse and report on for you - for a fee.  Otherwise, this is the tool
to gather data needed by the IBM support centre if you are raising it as a
problem possibly in need of a fix.

The we in this case is John and whoever else he involves. Just not me - for rootvg I am just the webmaster  :)

But, FYI - one of John's nicknames is "Mr. Performance".

Ah yes, the Royal "we" - blue of course - however, IBM have also been known to charge the odd fee; indeed, for my time when I was on staff analysing perfpmr data  ::)  It came to mind because I am working on such a report at the moment, in my now private capacity, and it's a nice way to get all the various stats and configs together around a problem.