Processing math: 66%

The Guru College

Home Network – Time Server

This is really a lot easier than it sounds. In essence, you have two of your computers talking to time sources, and then checking with each other to make sure they agree on what time it should be. If your internet connection goes down, or your hardware time source goes flaky, you keep a somewhat sane time on your home network until the problem is repaired. This also allows you to reboot your time servers, or do maintenance on them without disrupting the home network.

Running your own time servers has the added benefit of having the time stamps on all your logs match, and allows you to run your own time-sensitive services. Perhaps AFS or Kerberos? If you want to run an OpenDirectory or ActiveDirectory system, you’ll need time sync, as you’ll be using Kerberos.

So, cutting to the chase – here’s the config on the two Solaris Nevada time servers, vault and thor:

vault (/etc/inet/ntp.conf):

`This is really a lot easier than it sounds. In essence, you have two of your computers talking to time sources, and then checking with each other to make sure they agree on what time it should be. If your internet connection goes down, or your hardware time source goes flaky, you keep a somewhat sane time on your home network until the problem is repaired. This also allows you to reboot your time servers, or do maintenance on them without disrupting the home network.

Running your own time servers has the added benefit of having the time stamps on all your logs match, and allows you to run your own time-sensitive services. Perhaps AFS or Kerberos? If you want to run an OpenDirectory or ActiveDirectory system, you’ll need time sync, as you’ll be using Kerberos.

So, cutting to the chase – here’s the config on the two Solaris Nevada time servers, vault and thor:

vault (/etc/inet/ntp.conf):

`

thor (/etc/inet/ntp.conf):

``This is really a lot easier than it sounds. In essence, you have two of your computers talking to time sources, and then checking with each other to make sure they agree on what time it should be. If your internet connection goes down, or your hardware time source goes flaky, you keep a somewhat sane time on your home network until the problem is repaired. This also allows you to reboot your time servers, or do maintenance on them without disrupting the home network.

Running your own time servers has the added benefit of having the time stamps on all your logs match, and allows you to run your own time-sensitive services. Perhaps AFS or Kerberos? If you want to run an OpenDirectory or ActiveDirectory system, you’ll need time sync, as you’ll be using Kerberos.

So, cutting to the chase – here’s the config on the two Solaris Nevada time servers, vault and thor:

vault (/etc/inet/ntp.conf):

`This is really a lot easier than it sounds. In essence, you have two of your computers talking to time sources, and then checking with each other to make sure they agree on what time it should be. If your internet connection goes down, or your hardware time source goes flaky, you keep a somewhat sane time on your home network until the problem is repaired. This also allows you to reboot your time servers, or do maintenance on them without disrupting the home network.

Running your own time servers has the added benefit of having the time stamps on all your logs match, and allows you to run your own time-sensitive services. Perhaps AFS or Kerberos? If you want to run an OpenDirectory or ActiveDirectory system, you’ll need time sync, as you’ll be using Kerberos.

So, cutting to the chase – here’s the config on the two Solaris Nevada time servers, vault and thor:

vault (/etc/inet/ntp.conf):

`

thor (/etc/inet/ntp.conf):

``

Also be sure to have several different external time servers in the config files. At times, time servers go bad. The protocol is designed in such a way that if one source goes bad, it can easily fall back on the remaining.

Bandwidth Caps – The Sensible Version

I have a sensible, logical plan to deal with the issue of bandwidth capping. Charge 5/monthand0.10/100kbps/day. Importantly, let the end user select the maximum available bandwidth. Model the data using a 95% average model so the highest 5 minutes of use a day don’t count towards the billing cost, which allows for sudden spikes in traffic not driving the day’s cost up. The end result of this is that the user pays the bandwidth used. People who use more, on average, will pay more for it, but it will be clear what the prices can and will be. This addresses the problem the ISPs are facing: trying to get heavy users to pay “their fair share”. So, someone who only checks their email pays a pittance for service and the guy who wants to download 50 linux ISO’s gets to pay a little more for it. Better yet, when you go on vacation, and nobody is using the connection, you only get the $5 “service fee” for the month.

Someone who uses a full 1mbps connection 24 hours a day, every day, for a month pays 35forit,anda2mbpsconnectionwouldcost65. That may sound like a lot, but most people aren’t running at %100 capacity all the time. Looking at my own data use patterns, I’d be able to get a 10mbps connection for about 50 a month – considering that most weekdays we don’t use the internet very much. This destroys the old model of “pay N.NN a month to have access to a set limit of bandwidth – with no promise that the bandwidth is actually available when you need it”, and replaces it with a system that works.

And Still No Coffee

Wow. I forgot to drink coffee Saturday morning when we left to go to see JP, Nell and Kerian, and then again on Sunday. Well, I managed to get through the work day on Monday without any coffee (just a little sweet tea at lunch), and then all of Tuesday. It looks like I may get a chance to kick this habit once and for all. I’ve never made it this far without being downright miserable.

Here’s to hoping.

Home Network – Log Analysis

I’ve setup Splunk for my home network’s log analysis. Pretty sweet, if you ask me. It makes the watching and analysis of logs so much easier.

First – setup loghost and network logging. Remember – after these changes, you are going to have to restart syslogd. On each host sending in logs:

echo "*.err;kern.debug;daemon.notice;mail.crit @loghost" >> /etc/syslog.conf

And then, update the /etc/hosts files to have a loghost entry. This way, even if your local DNS server is down, syslogd can still find it’s log server.

On the Solaris 10 loghost:

svccfg -s svc:/system/system-log setprop config/log_from_remote = true<br /> svcadm restart svc:/system/system-log

That’s it. You will now have syslog traffic running to your loghost. Now, it gets really cool. Go to splunk.com and register to get the download link. They have packages for Solaris 8, 9, 10 (intel and x86) as well as Windows, Mac OS X, FreeBSD, Linux, etc. Unpack and install. The free license is good for up to 500 MB of uncompressed logs a day. I don’t think I can generate that much log data with the boxes at my house.

After you install the package, fire up http://loghost:8000 and go to town. The first thing to do is add a data source of /var/adm/messages , which is most likely where syslog is writing all the network log traffic to. It will load the log file, mine the data, and then keep an eye on it. All searchable, cross-referenced, and indexed for your viewing pleasure.

I’m not kidding. You too can watch your own logs with maybe an hour spent learning how to turn on network logging for your particular brand of *NIX, and then 10 minutes to install the package and configure it. And that’s being generous. It’s easy, and can save you a lot of pain down the road.

As a side note, you are going to want to make sure your time sync is correct on all your hosts, so the log event correlate properly. Perhaps now is time to setup a local time server?

Time Lapse

Bathtub IV from Keith Loutit on Vimeo.

Happy Mothers Day!

To my mother, the mother of my child, and to all mothers out there, everywhere: Happy Mothers Day! We sons love and cherish you! I don’t think there’s much of a better thing than a mother. Enjoy your day as much as you can!

Things He Will Always Know

It struck me the other day that ever since Qais was born, I’ve had an iPhone.

More to the point, multi-touch displays are a common thing, and he’ll never know a world where it’s just some cool technology demo that may or may not make it to the consumer market. When people bring up BetaMax versus VHS, he’ll know it as a legal ruling, not as a choice consumers had to make. He’ll probably never have much if any use for audio cassette tapes. I doubt that he will ever shoot on film, unless he’s making movies. And even then, possibly not. He’ll never use a Polaroid Instamatic camera. Branching out, the internet has always been available everywhere you go. There has always been a commercial space industry for tourists. Gay marriage has always been legal in some states.

There’s the Beloit Mindset List is made for graduating high school students that reminds all the rest of us just how old we are. Just trips me out really badly to think of what’s going to make it on this list 18 years from now that don’t occur to me. The one that hit me is #32 from this year’s list: There has always been Pearl Jam.. Honestly, this is all just blowing my mind.

Newer Posts | Home | Older Posts