monitor of specific processes

[pcov...@gmail.com]

I tried looking for what I want in a number of places with no luck, you can point me to it if you happen to know where I may find it,

I need to monitor the following and not sure how to go about it hence why I was looking,

we want to monitor the system and record for all the top process, cpu, users, dio and bio. I think that was it, and then hourly mail it. is it possible to do in one command procedure/output file. I have T4 running but it is unable to give us the detail needed.

we are having odbc timeouts to the system and it is causing the application to slow down to a crawl, the system is fine throughout this.

any thoughts?

thanks
Paul

[abrsvc]

I'm not 100% sure what you actually need here, but to the ODBC
question:

I would look at network traffic and possibly the collision rate rather
than the system involved here. Significant delays at the network
level are more likely to cause the ODBC problems. Do you have the
capability to monitor the network traffic?

Dan

[pcov...@gmail.com]

we are looking to see if there is one process that is causing our odbc timeouts. as far as network traffic. we are not seeing anything, the other morning it happened between 1 and 6 am. and eia0 showed virtually nothing hitting it... and there is no one working during those hours, just automated jobs. and it is sporadic... no set time or day!

[abrsvc]

On Feb 7, 12:59 pm, "pcovie...@gmail.com" <pcovie...@gmail.com> wrote:
> we are looking to see if there is one process that is causing our odbc timeouts.  as far as network traffic. we are not seeing anything, the other morning it happened between 1 and 6 am. and eia0 showed virtually nothing hitting it... and there is no one working during those hours, just automated jobs. and it is sporadic... no set time or day!

The question then becomes:

Is there a connection that exists that is timing out and disconnecting
or is a connection attempting to be established that is failing. In
the first case, an inactivity timer may be closing the connection. In
the second, it still points to a network connectivity issue rather
than a performance problem. I would attempt to log the network
activity on the receiving side using a packet trace utility. With the
ODBC failure, you should know the time and machine address. Search
the traces for that info and see what is happening. Chasing the
performacne of the machine here, I believe, will just waste time.

Dan

[pcov...@gmail.com]

we started wireshark and this is what we have found so far

But it looks like one side asks for a reset… and that’s the last request.

[Bob Gezelter]

On Feb 7, 11:59 am, "pcovie...@gmail.com" <pcovie...@gmail.com> wrote:

Paul,

T4 simply uses MONITOR/RECORD to capture the actual data. One can
manually (or in a custom batch job) use MONITOR/RECORD in a similar
manner as one chooses. (Been there, done that, particularly on older
systems).

- Bob Gezelter, http://www.rlgsc.com

[onedbguru]

ODBC from where to what?? Oracle? Rdb?

[pcov...@gmail.com]

the odbc connections is between sql servers and cache running on vms 8.3 1h1 and using the cache odbc driver on the servers.

and yes monitor/record but unless I'm missing something I have to run each individually and have a file for each to get what I want for output :-(

thanks
Paul

[hb]

On Feb 8, 7:46 pm, "pcovie...@gmail.com" <pcovie...@gmail.com> wrote:

> and yes monitor/record but unless I'm missing something I have to run each individually and have a file for each to get what I want for output :-(

Yes, monitor/record writes binary data to a file. But no, monitor can
combine the output for several classes. It seems you want the same
class but with different qualifiers, which works like this:

$ monitor/inter=1/end="+0 1::"/nodisplay/record process/topcpu,process/
topuser,process/topdio,process/topbio

The result is one monitor.dat file, with the data from all classes,
which you can display separately for example with

$ monitor/input process/topcpu

[pcov...@gmail.com]

yes exactly! thank you!!!!