This Old SPARC
The story of a geek and his Sun SPARCstations (and UltraSPARCs), and the trials, tribulations,
and small victories that go with the care and feeding of such a menagerie.
Sunday, August 27, 2000
I installed 'top' last night, from solarisfreeware.com. Top is great, because it gives you a "screen" of system information that updates in place. You can sort and filter the list of processes it displays, and keep track of free memory, CPU usage, etc. Here's an example screen:
last pid: 626; load averages: 0.36, 0.36, 0.32 10:49:03
51 processes: 50 sleeping, 1 on cpu
CPU states: 97.2% idle, 0.5% user, 1.6% kernel, 0.7% iowait, 0.0% swap
Memory: 128M real, 24M free, 42M swap in use, 133M swap free
PID USERNAME THR PRI NICE SIZE RES STATE TIME CPU COMMAND
411 thomasmi 14 25 0 32M 25M sleep 17:36 16.60% iexplorer
230 root 1 59 0 12M 8544K sleep 4:15 1.69% Xsun
514 thomasmi 1 48 0 2068K 1336K cpu0 5:55 0.27% top
367 thomasmi 1 48 0 3940K 2928K sleep 0:01 0.13% cmdtool
333 thomasmi 1 59 0 2924K 2136K sleep 0:03 0.04% olwm
165 root 8 53 0 2188K 1540K sleep 0:02 0.01% nscd
194 root 1 58 0 936K 496K sleep 0:02 0.01% utmpd
192 root 5 30 0 2300K 1688K sleep 0:01 0.00% vold
372 thomasmi 1 49 0 4036K 3040K sleep 0:01 0.00% perfmeter
217 root 5 22 0 2656K 1592K sleep 0:00 0.00% dmispd
128 daemon 3 23 0 2204K 1452K sleep 0:00 0.00% statd
315 root 1 32 0 3152K 2260K sleep 0:00 0.00% rpc.ttdbserverd
335 thomasmi 1 33 0 3816K 2148K sleep 0:00 0.00% olwmslave
218 root 4 33 0 3224K 2096K sleep 0:00 0.00% snmpXdmid
183 root 3 33 0 1148K 664K sleep 0:00 0.00% powerd
In a previous entry I referred to an 'awk' script I wrote that processes the output from /usr/ucb/ps to determine used/free memory. It turns out that the calculation isn't that accurate. Note that 'top' reports that I have about 24M free. Here's the output from my 'memfree' script:
679 /export/home/thomasmi>memfree
16.3593 %
However, 24M is really about 18.25% of memory (128M total). Oh well, I guess it's close enough for government work.
posted by Mike Thomas 8/27/2000 10:51:30 AM
Friday, August 25, 2000
Oh, I also tried something pretty cool today, since I had some time on my hands. I dialed into my company from my WinProxy computer (a Windows box) using the 56kbps modem in the box. This box acts as a gateway for the rest of my network. I then went to Sparky and telnet'ed to our development box at work. It worked like a charm! It shouldn't surprise me at all, and I'm surprised it took me this long to think of it. It will be much more convenient to work (or handle the occasional support call) from home occasionally, now that I can run a nice, big telnet session under Unix, rather than under the crappy 24-line telnet that comes with Windows. Next I'm going to try running an X application (like perfmon) over the dial-up connection, just to see how well (or badly) it works.
posted by Mike Thomas 8/25/2000 10:22:08 PM
It's been a few days now, and IE 4.0 has proven to be fairly stable, excepting a couple of lock-ups. I use it all the time now, even to post my blog entries.
I just received a remote for Sparky's monitor that I won from eBay. What a great monitor, but what a goofy idea -- "yeah, let's put all the controls on a remote!". Why? Remotes get lost, and besides, how often do I need to adjust the focus on my monitor from across the room? Anyway, all is well now. There's even a little slot at the bottom of the monitor fascia to hold the remote.
posted by Mike Thomas 8/25/2000 10:14:15 PM
Monday, August 21, 2000
OK, it seemed to work that time, and IE 4.0 has been running longer than IE 5.0 ever did. Maybe I'm out of the woods?
posted by Mike Thomas 8/21/2000 09:45:25 PM
The previous post worked, but the publish step didn't. It looks as though IE tried to create the FTP login dialog required to actually publish my comments, but didn't draw it right. It came out all screwy. I tried to resize it so I could fill it in, but when I did that it crashed. What a piece of crap!
I'll try to post this first, then publish.
posted by Mike Thomas 8/21/2000 09:42:41 PM
OK, I'm posting this using Internet Explorer 4.0. I figured maybe it was a little better tested than 5.0. So far so good, but 4.0's handling of CSS leaves a bit to be desired. Whatever, it works better than Netscape. I would like to have tried Netscape 6.0, but they don't have it for the Solaris platform. Damn.
posted by Mike Thomas 8/21/2000 09:37:14 PM
Web Browsing -- Not! After a couple of hours messing about with Internet Explorer for Solaris, I've given up on it. It locks up pretty reliably after about 10 minutes of use. If only it could work reliably, rather than breaking reliably. I've basically given up hope of doing any significant Web browsing on Sparky, for now. Netscape is too slow and stupid (e.g. if I resize the browser window, it round-trips to the server again and re-renders the page), and has crappy support for the things that make pages look good (CSS). IE does most things right, but can't seem to run for more than 10 minutes. Sigh. If only Opera for Solaris was available -- maybe then I'd find joy.
posted by Mike Thomas 8/21/2000 07:29:15 AM
Sunday, August 20, 2000
OK, third try now. What am I talking about? Well, Netscape is just too unusable to continue using, so I finally gave in and installed Internet Explorer for Solaris on Sparky. After what seemed to be a successful install, I was writing a blog about how much better IE is than Netscape, when IE locked up on me tight as a drum. I started again, and it locked up again. I started again, and it locked up just trying to render the blogger site. So I rebooted Sparky (horrors), and have started again. IE, in terms of usability and rendering, beats the pants off Netscape Navigator. However, if it's going to keep locking up on me, well that's another matter. I'm going to post this now, just to be safe.
OK, the post worked. Let me count the ways that IE is better than Navigator (under Solaris, at least):
- Important: I can actually post something from blogger, which I can't do reliably on Netscape.
- Better rendering
- Sites just look better. For example, the Post-entry page on blogger looks right
- Better fonts
- More flexible font sizing
- Better support for CSS
- Comes with better fonts
- I'm used to it.
Now if I can just prevent IE from crashing, I'll be in business. I'll keep you posted.
posted by Mike Thomas 8/20/2000 09:42:35 AM
Saturday, August 19, 2000
Fat software! I wrote a script tonight to add up all the memory used by the usual processes running on Sparky:
/usr/ucb/ps aux | awk '{tot += $4} END {print tot}'
It reported about 95%. I have 128M of RAM. This seems to suck. Netscape takes 25% all by itself. Time for more RAM?
posted by Mike Thomas 8/19/2000 10:15:04 PM
Tuesday, August 15, 2000
A good Solaris resource: SolarisGuide.com: The Unofficial Guide to the Sun Solaris Operating Environment.
posted by Mike Thomas 8/15/2000 07:28:35 AM
Monday, August 14, 2000
Sparky now has room to breath! He's got 64M more RAM, which makes for a total of 128M. That should hold him for a while. The install was painless, except for the weird order in which you're supposed to fill DRAM slots on a Sun. For tons of useful configuration details see docs.sun.com.
posted by Mike Thomas 8/14/2000 09:01:21 PM
Thursday, August 10, 2000
Sparky has a disturbing habit of being unable to see the internet if WinProxy is ever restarted on my proxy machine, such as when I reboot it. That's strange -- I thought "the network is the computer", and therefore, it seems that a given Sun box on Solaris should be pretty robust when it comes to networking stuff. My Windows clients can deal with this situation transparently. I have to reboot Sparky before he'll see the gateway again.
posted by Mike Thomas 8/10/2000 07:53:14 AM
Sunday, August 06, 2000
Sparky is back to life. It was simplicity itself. All I needed to do was reset the drive back to target 2, and then, in the EPROM do a "setenv boot-device=disk2". Now Sparky's lizard-brain looks for disk2 and the disk is disk2.
With Sparky back to life, it was now time to play with some things. First on the list: get an XWindow server to run on my Windows boxes. This would allow me to access Sparky through something other than telnet. Of course, there's not much to access on Sparky other than simple XClients like xload and xterm, but what the hell -- it would still be fun. So I searched around for an XWindow server, and found MI/X from MicroImages.
Installing MI/X was very simple. The install onto Sojourner (my laptop) took about 30 seconds. It wants to reboot your machine, but really doesn't have to (I wonder how many Windows products want to reboot, but don't have to?). Anyway, the only trick is that MI/X doesn't "bootstrap" your connection to the machine running XClients. You've got to run telnet to get that started. So, I telnet'ed into Sparky from Sojourner and exported Sojourner's display: "export DISPLAY=sojourner:0.0". Then I ran "xterm&". To my glee, I found an xterminal running under MI/X on Sojourner. now that xterm was running under MI/X, I could start other programs from X. I played around with different XClients for a while, and finally decided to run the big test: Netscape. This would be cool. Run Netscape on Sparky with its output redireced to Sojourner. Cool, but useless (after all I could run Netscape on Sojourner, too).
Anyway, it worked like a charm! Netscape actually looks better running under MI/X than it does on Sparky?!?. Part of it is the color depth, I'm sure (Sojourner is 24-bit color).
posted by Mike Thomas 8/6/2000 02:10:57 PM
Friday, August 04, 2000
Well, I found out yesterday (on SunHelp.org), having sent the money for the ZX graphics card, that the card won't work under Solaris 7. Damn. Screwed up again. Contacted the eBay seller for this item and begged him not to force me to buy it. He was a prince about it, and we agreed that he'd return my money, minus the eBay commission (they charge 2.5% of the winning bid as a transaction fee).
The guy that sold me Sparky sent me another case to help in repairing Sparky's lameness (he's missing a foot, and another is cracked). In order to make use of the new case, I had to swap all of Sparky's guts into the new case. I was a little intimidated by the operation at first, but Sun has done some great engineering on these machines. The operation went flawlessly, with the help of Sun's documentation site.
I did have one problem, however. Sparky couldn't seem to find his boot disk. He was looking for device 3, which he couldn't find. (Philosophical aside - Sparky isn't really sparky any more, if you consider that the EPROM that represents his long-term memory, or lizard-brain, is part of his motherboard, and the motherboard tends to go with the case. This is why Sparky couldn't find the disk). Anyway, I took the case apart again and set the drive's number to 3. Now the drive could be found, but the mount points couldn't be resolved to the filesystems that are defined as xxt2xxxx. I hunted around for a while in the EPROM to see if there was an environment setting to say what drive to boot from, but couldn't find anything.
So, for now, Sparky is only semi-conscious.
posted by Mike Thomas 8/4/2000 12:01:07 PM
Could be useful: How much RAM is enough?
posted by Mike Thomas 8/4/2000 08:11:08 AM
Thursday, August 03, 2000
I sent the money for Sparky's new graphics card today. I'll have to research how to get the card, and especially the driver(s), installed. Once that's done, I'll be rocking with true color. I'll be sending out more money for the 64M of RAM that I recently won, too.
I had won one auction for an SM51 processor card, and was well on my way to winning another one by the same seller, when the guy who sold me Sparky informed me that I can't put four processor cards in him. Doh!! Sparky will support four processors, but only by installing 2 x 2-processor cards. So, I nicely asked the seller if he'd let me off the hook. Luckily, he was cool about it, and freed me from my obligations. It would have been a real bummer to have to resell those items.
posted by Mike Thomas 8/3/2000 07:20:47 AM
|
