LinuxQuestions.org

LinuxQuestions.org (/questions/)
-   Linux - Server (https://www.linuxquestions.org/questions/linux-server-73/)
-   -   Performance troubleshooting tool (https://www.linuxquestions.org/questions/linux-server-73/performance-troubleshooting-tool-701079/)

amoralejo 01-30-2009 01:53 PM

Performance troubleshooting tool
 
Hi,

For our servers running RHEL4 we use HP Glance and Performance Agent to troubleshoot performance problems. Now, we are planning to migrate to RHEL5 and we are looking for an open source alternative (supported by RedHat if possible). We like glance and measureware but we have to pay for them. The requirements for the tool is to be able to find all relevant aspects in an easy way both global and for each process, including memory segments used, system calls, cpu usage, io activity per disk or volume and i/o performance, service time, etc... network statistics, information to a deeper level that the usual sar, iostat, vmstat, etc... on the other hand we'd like to be able to store this information locally for some time to be able to troubleshoot problems that may arise. I've found that with RHEL we have tools like oprofile, systemtap or frysk, but as far as I know are not easy to implement, don't provide all information in a fast and easy way.

What experience do you have with this kind of tools?, have you had good experiences with any specific product?

Best regards,

Alfredo

jstephens84 01-31-2009 09:50 AM

Quote:

Originally Posted by amoralejo (Post 3426593)
Hi,

For our servers running RHEL4 we use HP Glance and Performance Agent to troubleshoot performance problems. Now, we are planning to migrate to RHEL5 and we are looking for an open source alternative (supported by RedHat if possible). We like glance and measureware but we have to pay for them. The requirements for the tool is to be able to find all relevant aspects in an easy way both global and for each process, including memory segments used, system calls, cpu usage, io activity per disk or volume and i/o performance, service time, etc... network statistics, information to a deeper level that the usual sar, iostat, vmstat, etc... on the other hand we'd like to be able to store this information locally for some time to be able to troubleshoot problems that may arise. I've found that with RHEL we have tools like oprofile, systemtap or frysk, but as far as I know are not easy to implement, don't provide all information in a fast and easy way.

What experience do you have with this kind of tools?, have you had good experiences with any specific product?

Best regards,

Alfredo

have you guys given thought to say nagios or zenos core? they may do what you need and more.

amoralejo 01-31-2009 02:42 PM

That's not what we are looking for. We already have a monitoring tool that gather performance data. However, this kind of tools is not very usefull for complex problems because it provide only basic info as average resource usage, load, number of process, etc... For some complex problems we need more detailed info for each process, as memory areas for each process, what kind of system calls is using each process. We'd like to have a local tool to access and store this information locally.

Regards,

Alfredo

jstephens84 01-31-2009 04:44 PM

Then the best guess that I would say would be to script it. I can see if I have any of my old scripts laying around. But I used to have a couple that would obtain detailed information then send warnings out when specs seemed unusual. I did a quick search and could not find any alternatives for it.

amoralejo 02-01-2009 12:26 PM

After some research the best tool I've found is collectl. It can get most information I need and run as a service, saving the infor in a compressed file:

http://collectl.sourceforge.net/


You can use it together with gnuplot to create diagrams.

Another tool is collectd, however this is not exactly what I'm looking for.

Regards,

Alfredo


All times are GMT -5. The time now is 11:12 PM.