All proceeds from Ad Clicks goes to the author of this site.

 

Wednesday, November 11, 2009

1800 miles, 4 states...

1800 miles, 4 states, 2 hours on the tarmac, 1 hour job interview, 20 hour day, now I’m home.

Yesterday started at 4:30am with my alarm going off, after a fitful 3 hours of sleep, yes I’m a procrastinator was up to midnight packing the laptop, making sure the phone and the mp3 player was fully charged and had new stuff to listen too. Throw on clothes head off to the airport. And get on the plane to Memphis, Tennessee at 6am, as it seems with all planes I have been on lately they have been full, well I thought I was going to get lucky and have an empty seat next me, or I did until about 3 minutes before the plane’s door was closed and then the seat was at least filled, but all and all a pretty non-eventful flight.

In the intervening hour, mostly spent going between concourses and hitting the Starbucks standing in line, getting a Latte and a piece of pumpkin cake and then heading off to the next plane getting there in a few minutes before the 30 minute pre-flight boarding, flight was nice, but full but quite nice and smooth despite the remnants of hurricane Ida making Charlotte, NC very rainy and wet.

Landed met with the Agency person, and had a nice lunch of Red Robin, then off to the interview still about 30 minutes early after going through 2 security stations, and got to sit in the lobby for 30 minutes.

Interview went well I thought, they ran out questions to ask because this was now the 3rd time we had talked, it was only an hour, did I really sign up to fly 1800 miles and a 20 hour day for a 1 hour interview?, at least the agency paid for the flight. Went back to the Airport was about 4 hours early finally tracked down the right concourse and gate, my flight wasn’t even on the departure screens yet it was that early, had a decent Chinese food at a fast food place in the Airport, then proceeded to the gate and found an outlet so I could recharge my laptop and mp3 player that managed to power itself on and play during lunch and the interview using half its power.

About half an hour later still 2 hours before the flight they called all the passengers waiting for the 7:15 flight up to the gate, and told us our flight was delayed because of the weather in Atlanta, and wanted to move us to the earlier flight, which was really the 5:35pm flight that was now scheduled to leave at 6:40. Well I thought okay I don’t care since I can sit in any airport, I already had an hour layover in Atlanta anyway.

So after boarding the 6:40 flight which seemed to take much longer than usual because they had to put everyone from the earlier flight and the later flight onto the plane we leave the gate, taxied to the designated waiting zone, to sit on the tarmac, for about 30 minutes, then they turn off the no electronic device sign so we can listen to our mp3’s and laptop and even cell phone because we have to wait her 15minutes more before we can get into the official take off line, then 10 minutes later they turned the signs back on because we were going to get in line for takeoff but it was just tease because after the engine test the turned the signs back off because for some reason it would be an additional 30 minutes, well to make a long story short we actually got to the takeoff line at 7:55 pm, which was 40 minutes after the originally scheduled take off time of the later flight and 2 hours and 45 minutes after we boarded the plane.

To add insult to injury when we got into the air they announced that because the flight was so short about 1 hour and 2 minutes there would be no beverage service for coach how nice, strange I have had drink service during a 30 minute flight between Milwaukee and Chicago, we were held hostage in the plane for 3 and ½ hours and they couldn’t even be bothered to bring out the drink cart, at least they did offer cups of water to those that wanted it on the flight. I think it should be a rule that the time stuck on the tarmac should be included in the flight time calculations and they should be forced to provide beverages and snacks accordingly.

Well make it to Atlanta finally the gate staff which should have been apologetic seemed to care less that the 5:32 pm flight arrived 4 hours late at 9:40pm, and could be barely bothered to check the status of my flight to Milwaukee that was supposed to leave 9:42pm, which luckily delayed to 10:20pm so I got to run through the airport to reach my flight which of course was on a different concourse from where we landed.

As hard as it is to believe the I actually got an empty seat next to me on the final leg of the flight into Milwaukee and full drink service even thought the flight was only scheduled to last 28 minutes longer than the haha 1 hour and 2 minute flight from Charlotte to Atlanta. At 11:30pm I finally arrived back in Milwaukee 18 hours and 30 minutes after I left on my journey,

Home at last, now I just have to wait to see if I get the job.

Tuesday, October 13, 2009

Got dual cores/cpu or more? or maybe just more than one disk.

I came across Thomas Maier-Komor’s home page when I found this cool utility he wrote xjobs, it takes what xargs does and makes it multicore aware so that it starts x number of jobs to get things done faster, which is really cool to me. Before xjobs I found my self always writing little scripts like below, so I can use all my cores and may even take advantage data that is cached and have the operation complete faster.
for I in *.zip ; do   unzip $i & done 
But this doesn’t allow me to control how many processes get started off at any given time. This is where xjobs comes in. It basically does the same thing but starts the number of processes as the system has processors, but of course this is configurable. Since I didn’t want to repeatedly uncompress files to test this, I decided to throw the fact that my systems have multiple disks and file systems and look at how xjobs can help with IO spreading as well as just the CPU load.
The task that I am testing with is find all copies of “tar” I have in /usr and /opt, traditionally I just use find, but that really isn’t very efficient since it looks at one file system at a time. I ran all of these commands multiple times, to be sure the Solaris has cached all relative metadata so none of the commands should require physical IO to the disks. First on my Dual core AMD box, /opt is on my non-root pool composed of 4x 512GB sata disks, and /usr is on my root pool made of a single 250GB sata disk.
Xjobs arguments used –j sets the number of jobs to use to complete the task and –v 0 tells it to be quient about process starting and ending.
To make find, xjobs friendly I wrote a small wrapper script

qfind:
#!/usr/bin/bash
/usr/bin/find $2 -name $1 2> /dev/null


The actual tests:
#xjobs using two processes.
$ time echo -e "/usr \\n/opt \n" | xjobs -j 2 -v 0 ~jamesd/xjobs/qfind tar
/opt/csw/bin/tar
/usr/bin/tar
/usr/gnu/bin/tar
/usr/sbin/tar

real 0m2.159s
user 0m0.389s
sys 0m2.838s

#xjobs using just one process
$ time echo -e "/usr \\n/opt \n" | xjobs -j 1 -v 0 ~jamesd/xjobs/qfind tar
/usr/bin/tar
/usr/gnu/bin/tar
/usr/sbin/tar
/opt/csw/bin/tar

real 0m2.825s
user 0m0.375s
sys 0m2.443s


#the old fashion way, using find
$ time /usr/bin/find /usr /opt -name 'tar' 2> /dev/null
/usr/bin/tar
/usr/gnu/bin/tar
/usr/sbin/tar
/opt/csw/bin/tar

real 0m3.006s
user 0m0.365s
sys 0m2.633s


Now what surprised me the most, is that old fashioned find was slower than even xjobs when I limited it to just one process. Even with the extra work to spawn the xjob with one thread is a lot greater because it has to echo the two directories into xjobs and has to setup a pipe as well, and then run a script two times that spawns two copies of bash where find is just one command doing what it was designed to do.

Okay now if you have only one cpu but multiple drives does his help you? Let’s see. For this test I use my Blade 1500, single cpu, and 1 ide drive formatted with UFS for /usr and /opt on a zpool raidz composed of 3x 18GB 10k rpm drives.

frankenstein:xjobs-20091012$ time /usr/bin/find /usr /opt -name 'tar' 2> /dev/null\ /usr/bin/tar
/usr/sbin/tar
/usr/sfw/share/doc/ant/manual/api/org/apache/tools/tar
/opt/csw/bin/tar

real 0m5.827s
user 0m0.729s
sys 0m3.823s

frankenstein:xjobs-20091012$ time echo -e "/usr\n/opt" | /usr/local/bin/xjobs -v 0 -j 1 ./qfind tar
/usr/bin/tar
/usr/sbin/tar
/usr/sfw/share/doc/ant/manual/api/org/apache/tools/tar
/opt/csw/bin/tar

real 0m5.969s
user 0m0.736s
sys 0m4.034s

frankenstein:xjobs-20091012$ time echo -e "/usr\n/opt" | /usr/local/bin/xjobs -v 0 -j 2 ./qfind tar
/opt/csw/bin/tar
/usr/bin/tar
/usr/sbin/tar
/usr/sfw/share/doc/ant/manual/api/org/apache/tools/tar

real 0m5.264s
user 0m0.726s
sys 0m3.604s


The results show that xjobs was about .6 second or about a 10% gain by using xjobs, Since we didn’t have to access physical disks for any of these tests I think we can assume the results would be more impressive if the cache was cold.

Check out Thomas’s home page he has some other cool utilities, like multi-threaded tar patch and sysstat.

Monday, October 12, 2009

bored?

watch Oracle Open World live

Wednesday, September 16, 2009

ZFS ARC summary in web language

Lately I have been working on system monitoring scripts and decided to take Ben Rockwood’s ZFS ARC summary script and convert it to output in html format. This is just the first step I will be working on a configuration file to use MRTG to track this information s well. You can get the current release of this script at zfs_status_web.txt


Sample output, the script looks nicer because it doesn't conflict with my blog's CSS.






System Memory
Physical RAM6006 MB
Free Memory1062 MB
LotsFree93 MB






ARC Size
Current Size2115 MB (arcsize)
Target Size (Adaptive)3298 MB (c)
Min Size (Hard Limit)622 MB (zfs_arc_min)
Max Size (Hard Limit:4982 MB (zfs_arc_max)




ARC Size Breakdown
Most Recently Used Cache Size57%  1910 MB (p)
Most Frequently Used Cache Size42%  1387 MB (c-p)

Friday, September 11, 2009

Tracking down sunray's on ciscos

After installing mrtg/rrdtool on OpenSolaris Tracking down I wanted to see how much bandwidth my Sun Rays were using so I needed to know which port the sunray were plugged into on my Ciscos

Normally the last 2 lines is enough but ping is necessary at least on my Cisco 2950's and sunray's


ping [ipaddress]
show arp | include [ipaddress]
show mac-address-table address [mac address]


based on information I found at: http://www.velocityreviews.com/forums/t37560-how-to-track-down-whos-on-what-port-on-an-ios-6509.html


and for those that are interested here is the bandwidth usage of the sunray, its sitting idle and it uses very little bandwidth.

Monday, August 17, 2009

Happy B-Day to me!

It's my Birthday, wasn't it just my birthday 365 days ago? sure feels like it. It's going okay, got news that I have a 2nd interview lined up for a job I'm trying for. So seems like a good day so far.

Tuesday, August 04, 2009

Are you old enough to get the humor in this?

My friend Melanie posted this really funny skit, on her blog check it out.

Another funny story you don't have to old to get it, just married.

Interesting Links and Tools

Unix isn’t going away, and 3 that it may.
Found it really interesting that 1/3 of all server sales by dollar are Unix Servers.

Freebase Parallax
Cool search engine that changes the way you search from one topic to one page, a more to a more relationship type search. watch the 10 minute video on the home page.

VirtualBox 3.02 released

One of the best free virtualization tools, new version, supports SMP in virtuals now and many other features, Virtual box keeps getting better and better.

FoxTab
One of the coolest Firefox plugins gives you a 3D w to choose which tab you want, works great when you have browser windows open and each has more than 10 tabs. Gives you a way to show off that fancy 3D video card, with out loading a game.


Sync Back
Need to backup your home windows boxes? Syncback is a great way to do it, backs up to directories, windows shares, even my favorite FTP

WinDirStat
Gives you a way to find all those large that you know they are somewhere, but not sure where, and allows you to see what is really using all your hard disk space.