7

Is the following an appropriate question to ask here? if not, where should it go?

At my place of work there have been consistent though sporadic network issues for a long time, a couple of years, though in the last 6 months the troubles are increasing, sometimes to the point where my home 16mbps-down/768kbit-up cable connection is outperforming our work 100mbps connection.

I administer a handful of Win2003 servers but my network admin activities are limited to routing cables in the wiring room to free ports. I've brought the issue up with the next level of IT support, who commiserate but tell me there's nothing they can do because the switches are under control of central IT, and their pleas to have central IT look into things have fallen on deaf ears, or at least unresponsive ones.

We have access to all of the physical infrastructure in our building, so I could, for example, put in a passive something or other in-line for monitoring or benchmarking (provided there's a non-interference guarantee).

We're pretty much a 100% Microsoft shop, with authentication etc. handled by Active Directory, though there are a scattering of linux servers and VMs here and there doing odd jobs.

What can I do from the bottom of the chain to prove or disprove there really is a problem? and assuming there is one, localise it? (on 2nd floor, between machine X and external uplink, etc.)

What do you, as a top of the chain network administrator, need to hear to convince you there is something to investigate?

3 Answers 3

10

Your question is appropriate, except that it's not :)
There are two major issues:

  1. You're not the system administrator
    More specifically you're not the network administrator - that seems to be the role of this mysterious "Central IT". It's their system, and neither you nor your colleagues seem to have access to troubleshoot and fix it.

  2. Nobody on Server Fault can diagnose your network problems for you.
    We're not at your site, and we have no access to your infrastructure.

In order for this question to be a good fit you would need to be the netadmin, and you would need to do 90% of the troubleshooting yourself.
After doing that troubleshooting you probably wouldn't need us, because you would have a pile of evidence pointing at a problem you can fix (flaky switch, overloaded infrastructure, shoddy network uplink).


The best tips I can offer you are:

Determine if the problem is internal, or if it's limited to external sites.

If the problem is internal you will need access to your switches to do anything (at a minimum you need to inspect their traffic & CPU utilization). Continuous pings that traverse different portions of your switch stack may help isolate the problem further (and are something you can do without switch access) but you'd have to present these results to "Central IT" and convince them it's a real problem.

If the problem is limited to external sites perform continuous pings to a known-stable site and document latency spikes, periods of high packet loss, etc. - To really troubleshoot you'll probably need access to your edge routers to log interface errors as well, but again the continuous ping is something you can do with no special access.

Second, take your evidence from the above and whack "Central IT" upside the head with it. Tell your boss. Tell your boss's boss. Explain that the issues are trashing company productivity.
Visit The Workplace for some ideas on how to fight the political battles required to get your problem fixed (without torching all your bridges in the process).

1
  • thank you. I'll see what I can uncover with continuous ping, and especially thank you for The Workplace, a new SE site for me. You're correct that the real issue here is politcal/social, not technical. I guess I've been focussing on the technical because I'm good at it, and feel like I've a hope of moving things on that level, whereas the political... well after almost 15 years here there's a reason I'm still not a manager. :) Nov 29, 2012 at 19:00
6

You're in a bit of a gray area... We normally require that you have administrative privilege to the systems you're diagnosing. You have access to the servers, but it sounds like you don't have access to the network gear.

Unfortunately your first line of troubleshooting will be "log into the switches, find out what their error counts and such diagnostic counters show". Failing that, you'll probably have to setup something like smokeping between as many computers as you reasonable can, and see what it has to say. It will be much more time consuming and less accurate however; at least you'd have a quantitative baseline to prove the existence of a problem.

1
  • Thanks Chris. I'll look into smokeping. At this point any data will be useful. Nov 29, 2012 at 19:01
1

As has already been said, in this context you're a mere end user, which is a problem for us.

As for the issue itself, the problem, at least from your perspective, is more human than technological because the humans are blocking a technological fix.

I can see that your organisation has a number of IT areas and you're not all playing nicely together. However, there must be someone higher up to which you all answer. What you need to do is to present a business case for resolving the issues, with as much data as you can collect, and ask that it be investigated by all areas, each working on their specialised areas but without the exclusionist attitude that appears to be prevalent.

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .