I have a client for whom I have setup a cluster.Which is basically a few Virtual Machines running and the applications running in them which are accessible on internet.
He has asked me to send him a weekly report of this work.
I am sys admin guy who understands ssh,telnet,ftp,tftp,TCP
I am not able to understand what should I write in report.Because all the servers are perfectly running and applications are also running on top of them and I am done with this.So basically from my part I do not have any ssh or ftp to write in a report like this.Can some one give me a link if there is some sample report that I should send.I am not able to understand what do I need to Google for the same.

If you are still building the cluster then the report should list milestones toward the finish of the project, the work that was performed that week, problems encountered, problems solved, and an estimate on time to completion.

If the cluster is running then the report should list performance items, problems encountered, problems resolved, problems not resolved and the plan to deal with them, and a one line analysis of the suitability of the cluster for the workload. This weekly report lends itself to you being imaginative. Just think about what you would want to know if you had some critical asset that you didn't understand. What would you want to know? This kind of report can help the client feel more comfortable when the machine is working and it can provide a conduit for bad news when issues arise.

The performance report can be created by combining the results of various analytical tools such as sar and log file analyzers. You should be able to automate 90% of the report creation and add a small interpretation to each of the analysis sections. The whole thing could end up taking a half hour a week once you get your report building script running.

Do NOT use the report to "baffle them with "bull****". The manager probably has two goals for this report. One is to feel comfortable about this critical asset. The other is to provide traceability when problems occur. The manager wants to cover his *** when things go wrong. So, make sure that you report problems and potential problems as soon as you become aware of them and talk to the manager about them. That way the manager can be proactive and there is less justification to shift the blame to you when something bad happens. Naturally if this client is a manager in a large corporation then when something goes wrong you will be blamed BUT you will be in a more defensible position. So keep a copy of these reports for yourself.

I'll have to take a different approach. If your client has asked you to send a report and you don't know what to send: ask the client.

Do not give a report that you would want. Do not give a report that you think they want. Give them a report that they want.

Let me put it this way, if the client approached you and said, "build me a cluster." What do you do next? You have to ask more questions. What kind of cluster? How many machines? What kind of networking are we dealing with? Ethernet? Fiber? Token Ring??

The responsibility is on you to get all the information you need to complete the job. This is no different. If their request for a report is ambiguous, you have to ask more questions.

1 members found this post helpful.

That definitely is a different approach. My approach works if the client is not technical at all. Your approach works if the client is very technical.

Either way (technical or not), the client is the one that writes the check. It doesn't matter what kind of service agreement or contract that was signed, the client can always refuse to pay if they don't like what they get. Depending on the amount, it may not be practical to go after them (time and/or money-wise).

If you (as the employee/contractor) get the requirements in black-and-white, and meet those requirements, the risk of getting stiffed is enormously reduced.

But back to the technical/non-technical thing, I'd have to say my gut feeling would be that these folks are technically savvy. They wanted a cluster, which says "technical" to me. Not only that, they wanted a "virtual" cluster, which screams "technical" to me. Whether the OP's contact person with the client is technically savvy or not, someone in that organization is. And that is one of the questions that needs to be asked. If the contact person is not the technically savvy person, find out who is (because they're likely to be the ones managing/maintaining this project after completion), and determine what information they need--whether it be a project-build log, an automated script to monitor the health of the cluster after completion, or something else.

But that's just my take on it.

I think this is good thread even for newbies to system administration I would request people to share their thought as how it works.

I read all the replies. The thing is I am new to all such things and I have set up the cluster and every thing running and in action.
I understand that asking client would be best thing in this case.
But currently this is not the situation with me,

Ok this part is already not in picture as every thing is working perfectly.
Ok this is a good suggestion but can you tell how do you do a one line analysis of suitability of the cluster for the workload.
Can you give any sample report or some link where I can understand.
My creativity comes to a standstill for this part.
Please suggest some tools I never heard any such tools.Before asking this question here I tried to search on internet but I am not clear what should I search for this topic.
That is the reason I asked this question here.


Correct I want to be on safe side and the reports which I made are snapshots of server installations or consist of commands like
sudo apt-get or
some other non sense Unix stuff I did but how do I present it that I am not clear.
I do not want to bluff but the problem is I do not know how to represent in words what I did.I can spend hours on internet reading tutorials and doing things but I am clueless when it comes to mention that
I created a Xen environment and xen had crashed to which I had no clue and I searching such issues here and there.
So the reports which I have made till now I feel are not more than a piece of junk or some traces of log files or a script or two here and there.

I used to occasionally create these kinds of reports for my managers. I used the same tools that I was already using in my daily system administration routine to monitor system resources, user workload, application workload, etc. So you should consider doing this kind of research as part of your daily routine as well.

On Solaris machines I had sar and BMC Patrol installed. sar is a set of utilities that are included in the Linux sysstat package. I cannot write a tutorial on how to use it. Just install it and configure it to record data hourly. Then you can use it interactively or in a batch job to create resource usage reports. You can use this data to create graphs. The horizontal axis is time, the vertical axis is percent of resource usage, different color lines represent memory, cpu, network traffic, disk i/o, etc. You should probably run sar on the host machine as well as on the each of the virtual machines.

You should know who is using the computer and how they use it. Does the client have a database administrator who creates new scripts? Does the accounting department max out the capacity of the machine at the end of the week/month/quarter? Your report could include a judgment about the ability of the machine to handle the workload on an hourly, daily, weekly, and peak use basis.

I haven't had to do this kind of work on Linux machines. I have done it on Solaris and VMS machines. Either you will have to find the data analysis tools yourself or someone else will have to provide the names of these tools that are available on Linux.

And in that regard it would help if you told us your Linux distribution(s). Different distributions have different packages available. As far as I know all of the Linux distributions include sysstat so install that and configure sar to collect data hourly. That is certainly your first step.

There are log file analyzers to determine who accessed the machine or applications on the machine and when they did that. There are log file analyzers that look for security concerns such as detecting when someone else is scanning your network ports or when someone has too many failed login attempts.

A one (or two) sentence summary might look something like this:
The cluster has ample resources to handle the peak workload. The fault tolerant hardware helps to ensure that the cluster will continue to provide service in the event of a hardware failure.

This report is really just a summary of the things that you should already be checking to ensure that the computer performs its mission.

Thanks for the detailed reply.
Host is Ubuntu and one Guest is CentOS rest all are Ubuntu.