Filed under: OS Updates, Features, Linux, Open Source, Unix
OpenSolaris 2008.05, and other places the sun don't shine
by Kristin Shoemaker May 15th 2008 at 2:00PM
Way back in the dark ages of 1993, we were introduced to this thing called email. Email in the olden days was not like email now. All the packets traveled uphill no matter where they were going, and usually there was a good three or four feet of snow on the internet backbone. We used these big hulking things called VAX/VMS nodes that were attached to some pretty sweet fourteen inch monochrome VT 100 terminals. There were also these machines that ran something called UNIX, which sounded to us like something that should have been found in the college health center, not the computer lab.Eons passed, and things changed. Though there were many more email packets flying around, plate tectonics had changed the course of things so that now they traveled downhill, really really fast. The internet backbone became a series of tubes. All the VT 100 terminals banded together and created an archipelago in the South Pacific. And UNIX...
UNIX evolved. Mutated. It trickled down into various UNIX brands and distributions. There were things like BSD UNIX, HP-UX, and AT&T Bell Labs UNIX. There were other branches, too, rogue sprouts on the evolutionary tree: FreeBSD, and OpenBSD. Slightly alien but vaguely reminiscent life forms injected their DNA into the gene pool: Linux, and this weird little UNIX-esque animal called Solaris.
Sun recently let Solaris go open source. OpenSolaris is more a traditional UNIX environment than a Linux type environment, but the appeal of taking a peek at the 2008.05 OpenSolaris release was too great for us to resist. The folks at OpenSolaris knew this, and baked some goodies into the OS that no Linux user could refuse.
We were given a no-strings attached liveCD, so our Linux install would never know we cheated. We had a bash shell, and the GNOME desktop environment, so our eye candy and commands would feel familiar and easy.
Sometimes, though, evolution goes horribly, horribly wrong.
We eagerly booted into our OpenSolaris environment on our Athlon64 X2 system. We selected our keyboard layout and language, and OpenSolaris graciously presented us with the X server and GNOME. It was pretty, and we could even activate Compiz, of all things. Yes, it was very pretty. And for a reason we'll discuss a little later, we thought it would be really beneficial to stop what we were doing with the liveCD, and reboot.
OpenSolaris had other ideas. It just didn't want to tell us what those other ideas were.
GRUB would launch, and we could select the standard OpenSolaris boot option. The splash screen showed its mug, and we'd be dumped on to a screen with the standard trademark and legal disclaimers. And the cursor would blink, and blink, and blink. We waited. Patiently.
But it was clear that the optical drive was no longer reading, and the hard drive was inactive. Keyboard presses did nothing. Trying the whole process again with a text console GRUB selection played out the same way.
Certain things disturb us much more than they probably should. Humanoid puppets, hyper-intelligent birds, and when our computers behave irrationally for no apparent reason are the big three. OpenSolaris had thrown down the first gauntlet, and by honor, we could not walk away. We had to know not only why it booted once, but why it didn't seem to want to again.
The condensed version of the next 21,600 seconds (six hours, for those not wanting to do the math) went something like this:
It's a 64-bit versus 32-bit problem. We'll edit our GRUB boot lines so we boot into 64-bit no matter what.
It's a 64-bit versus 32-bit problem. We'll edit our GRUB boot lines so we boot into 32-bit no matter what, then.
Does this piece of crap have an issue with SATA drives? How can it? That makes no sense!
Take out a stick of RAM? Are they kidding? This is based on Solaris. How can it freak at 2 gigs of RAM?
Let's disable PnP in the BIOS. Oh. It already is.
Let's mess with our memory timing. Reduce by half, they say. We laugh in the face of unstable systems.
We have got to be crazy. We've got to. Why is it so damn important if this thing ever boots again?
Why is our IDE controller on in the BIOS anyway? We don't have any IDE drives. Not that it affects anything anyway.
Let's try every ACPI setting there is. Twice. Let's disable ACPI APIC. Let's enable it again.
Who would realistically ever go through this to boot an operating system? What is wrong with us?
Oh, hey, there is a switch for kernel debugging and a verbose switch in GRUB! Edit kernel line with -k -v and boot.
At the six hour mark, we approached the solution:
ehci is having an issue. That's... USB stuff? Why does it say it is ignoring the error? Duh. Obviously it isn't.
Hey, BIOS! It's us again. Bye, auto legacy support.
That didn't work. Do we have any USB devices even attached? The printer. The printer isn't on. Or plugged in.
Let's boot with -k -v again. There, it's talking about the keyboard, and then ehci freezes--
They have got to be kidding. Seriously. Our mouse? It is a USB mouse. Fine. We'll put a PS/2 adapter on it. If this works, we will renounce the idea that technology operates on any sort of logic. Let's reboot 'er.
Dear readers, all that technology you see on the desk in front of you today? Apparently, it is a collection of happy accidents. Putting a PS/2 adapter on our mouse allowed us to reliably boot into OpenSolaris.
This confounds us on a few levels. Solaris is a UNIX derivative. UNIX is (at heart) a command line that has the ability to run a GUI. It doesn't have to. We could have booted without a mouse. But booting with a mouse that wasn't necessarily fully recognized as such is apparently disastrous. Is it an input device issue? A USB issue?
Linux, which is also at heart a command line that is able to run a GUI, does on occasion have issues with various mouse types. USB support on Linux has also been historically buggy, though it is much better than it was even two years ago. However, we've never had a Linux machine hang indefinitely because the pointer device was somewhere the OS didn't want it. We've had plenty of experiences with non-working mice.
Non-working mice are a lot easier to troubleshoot than non-working systems.
We tried out a Zen outlook (we've learned so much through this agonizing enlightening process!) and began playing with OpenSolaris in earnest.
Remember the history of email we touched on at the beginning of this post? Slow, uphill both ways? Snow on the backbone? Migraine inducing VT100 terminals? They are all still more effective than trying to connect an OpenSolaris machine to the internet.
We admit that Linux has a long way to go with wireless support. However, we can not remember the last time we had a modern, full-bodied distribution have any real issues with a wired connection. The most we might have to do in Linux is activate the appropriate device, or point it at the right gateway.
OpenSolaris does have this nice Device Driver Utility. Utility is a bit of a misnomer. Unless you write device drivers, it doesn't really do much, except tell where on the system the device driver is located and whether the driver is supported and working.
Our ethernet connection is located at nge0 (as opposed to the Linux eth0). Good to know. Our driver is working for the ethernet card. Awesome. Er, so why can't we get online?
We can't reconfigure our card. OpenSolaris knows best. It uses this little Network Auto-Magic daemon. It thinks it is connected. It clearly isn't. Disabling the Magic is just different enough for a seasoned Linux user to feel a bit like they're in a bizarro universe, but at least the command line concepts between UNIX and Linux aren't horribly different.
On a liveCD such as OpenSolaris, manually configuring the ethernet device isn't a walk in the park either. There's that whole root password thing. Where it seemed we should be able to manually enter nameserver information, and our gateway, and all sorts of things, it didn't let us. We are not root, or remotely worthy.
We finally dug up the "pfexec su" command, and tried to get online that way. We could add our nameservers, and gateway, and we could activate it all, now that we were disguised as root. Apparently, though, OpenSolaris has a problem with our router, and the way it assigns various addresses. We may be activated. We are not connected. Where is VAX/VMS when you need it?
We'd just spent six hours getting OpenSolaris to boot. To hell with the internet connection. We are determined to the point of amazing stupidity, but even stupidity has limits. At least ours does.
Don't get us entirely wrong, there are some nice things about OpenSolaris. It is very light on pre-installed applications. There are the applications that come with GNOME, like gedit, some games, and system administration tools. There are a few multimedia apps, GIMP, and gtkam. That's about it. It's a nice way to build a system that does exactly what you'd like. (We are puzzled by the inclusion of Compiz, frankly, because it's a feature that doesn't contribute a whole lot of function to the desktop. It seems illogical to be there by default given what else is and isn't included.)
Despite our USB mouse woes, flash drives mount automatically and work as expected. Our audio drivers exist and supposedly function, but they've gone to ground with our network connection. So system sounds won't play through GNOME's esd. Because there are some gstreamer plugins that require installation via the packaging system, we couldn't try out any media files.
If we were hardcore software developers with a keen interest in UNIX-type environments, we'd go ahead and install OpenSolaris on our computers. We'd just rely heavily on the ol' VAX/VMS set up and the cutting edge Mosaic browser for our networking needs.
For the desktop Linux user who is simply curious about UNIX derivatives, we'd have to advise taking a pass on this release, anyway. We'd probably tell you to check out OpenBSD, or FreeBSD, for a shot at a more successful change of pace.
OpenSolaris had other ideas. It just didn't want to tell us what those other ideas were.
GRUB would launch, and we could select the standard OpenSolaris boot option. The splash screen showed its mug, and we'd be dumped on to a screen with the standard trademark and legal disclaimers. And the cursor would blink, and blink, and blink. We waited. Patiently.
But it was clear that the optical drive was no longer reading, and the hard drive was inactive. Keyboard presses did nothing. Trying the whole process again with a text console GRUB selection played out the same way.
Certain things disturb us much more than they probably should. Humanoid puppets, hyper-intelligent birds, and when our computers behave irrationally for no apparent reason are the big three. OpenSolaris had thrown down the first gauntlet, and by honor, we could not walk away. We had to know not only why it booted once, but why it didn't seem to want to again.
The condensed version of the next 21,600 seconds (six hours, for those not wanting to do the math) went something like this:
It's a 64-bit versus 32-bit problem. We'll edit our GRUB boot lines so we boot into 64-bit no matter what.
It's a 64-bit versus 32-bit problem. We'll edit our GRUB boot lines so we boot into 32-bit no matter what, then.
Does this piece of crap have an issue with SATA drives? How can it? That makes no sense!
Take out a stick of RAM? Are they kidding? This is based on Solaris. How can it freak at 2 gigs of RAM?
Let's disable PnP in the BIOS. Oh. It already is.
Let's mess with our memory timing. Reduce by half, they say. We laugh in the face of unstable systems.
We have got to be crazy. We've got to. Why is it so damn important if this thing ever boots again?
Why is our IDE controller on in the BIOS anyway? We don't have any IDE drives. Not that it affects anything anyway.
Let's try every ACPI setting there is. Twice. Let's disable ACPI APIC. Let's enable it again.
Who would realistically ever go through this to boot an operating system? What is wrong with us?
Oh, hey, there is a switch for kernel debugging and a verbose switch in GRUB! Edit kernel line with -k -v and boot.
At the six hour mark, we approached the solution:
ehci is having an issue. That's... USB stuff? Why does it say it is ignoring the error? Duh. Obviously it isn't.
Hey, BIOS! It's us again. Bye, auto legacy support.
That didn't work. Do we have any USB devices even attached? The printer. The printer isn't on. Or plugged in.
Let's boot with -k -v again. There, it's talking about the keyboard, and then ehci freezes--
They have got to be kidding. Seriously. Our mouse? It is a USB mouse. Fine. We'll put a PS/2 adapter on it. If this works, we will renounce the idea that technology operates on any sort of logic. Let's reboot 'er.
Dear readers, all that technology you see on the desk in front of you today? Apparently, it is a collection of happy accidents. Putting a PS/2 adapter on our mouse allowed us to reliably boot into OpenSolaris.
This confounds us on a few levels. Solaris is a UNIX derivative. UNIX is (at heart) a command line that has the ability to run a GUI. It doesn't have to. We could have booted without a mouse. But booting with a mouse that wasn't necessarily fully recognized as such is apparently disastrous. Is it an input device issue? A USB issue?
Linux, which is also at heart a command line that is able to run a GUI, does on occasion have issues with various mouse types. USB support on Linux has also been historically buggy, though it is much better than it was even two years ago. However, we've never had a Linux machine hang indefinitely because the pointer device was somewhere the OS didn't want it. We've had plenty of experiences with non-working mice.
Non-working mice are a lot easier to troubleshoot than non-working systems.
We tried out a Zen outlook (we've learned so much through this agonizing enlightening process!) and began playing with OpenSolaris in earnest.
Remember the history of email we touched on at the beginning of this post? Slow, uphill both ways? Snow on the backbone? Migraine inducing VT100 terminals? They are all still more effective than trying to connect an OpenSolaris machine to the internet.
We admit that Linux has a long way to go with wireless support. However, we can not remember the last time we had a modern, full-bodied distribution have any real issues with a wired connection. The most we might have to do in Linux is activate the appropriate device, or point it at the right gateway.
OpenSolaris does have this nice Device Driver Utility. Utility is a bit of a misnomer. Unless you write device drivers, it doesn't really do much, except tell where on the system the device driver is located and whether the driver is supported and working.
Our ethernet connection is located at nge0 (as opposed to the Linux eth0). Good to know. Our driver is working for the ethernet card. Awesome. Er, so why can't we get online?
We can't reconfigure our card. OpenSolaris knows best. It uses this little Network Auto-Magic daemon. It thinks it is connected. It clearly isn't. Disabling the Magic is just different enough for a seasoned Linux user to feel a bit like they're in a bizarro universe, but at least the command line concepts between UNIX and Linux aren't horribly different.
On a liveCD such as OpenSolaris, manually configuring the ethernet device isn't a walk in the park either. There's that whole root password thing. Where it seemed we should be able to manually enter nameserver information, and our gateway, and all sorts of things, it didn't let us. We are not root, or remotely worthy.
We finally dug up the "pfexec su" command, and tried to get online that way. We could add our nameservers, and gateway, and we could activate it all, now that we were disguised as root. Apparently, though, OpenSolaris has a problem with our router, and the way it assigns various addresses. We may be activated. We are not connected. Where is VAX/VMS when you need it?
We'd just spent six hours getting OpenSolaris to boot. To hell with the internet connection. We are determined to the point of amazing stupidity, but even stupidity has limits. At least ours does.
Don't get us entirely wrong, there are some nice things about OpenSolaris. It is very light on pre-installed applications. There are the applications that come with GNOME, like gedit, some games, and system administration tools. There are a few multimedia apps, GIMP, and gtkam. That's about it. It's a nice way to build a system that does exactly what you'd like. (We are puzzled by the inclusion of Compiz, frankly, because it's a feature that doesn't contribute a whole lot of function to the desktop. It seems illogical to be there by default given what else is and isn't included.)
Despite our USB mouse woes, flash drives mount automatically and work as expected. Our audio drivers exist and supposedly function, but they've gone to ground with our network connection. So system sounds won't play through GNOME's esd. Because there are some gstreamer plugins that require installation via the packaging system, we couldn't try out any media files.
If we were hardcore software developers with a keen interest in UNIX-type environments, we'd go ahead and install OpenSolaris on our computers. We'd just rely heavily on the ol' VAX/VMS set up and the cutting edge Mosaic browser for our networking needs.
For the desktop Linux user who is simply curious about UNIX derivatives, we'd have to advise taking a pass on this release, anyway. We'd probably tell you to check out OpenBSD, or FreeBSD, for a shot at a more successful change of pace.
Get a WordPress.com Blog
With Halloween fast approaching, it's a great time to get in some practice defending your territory against zombies. In Graveyard Shift, you take aim at zombies and other creepy-crawlies, blasting them into splatters of cartoony green guts. It's a casual first-person shooter, and it's very easy to get the hang of - use the mouse to aim, click to fire. Graveyard Shift has at least 15 levels, and it might even have some secret stages I haven't unlocked yet.
They key to getting good at Graveyard Shift is learning to use ...
Reader Comments (Page 1 of 2)
Jonathan Harford said 2:14PM on 5-15-2008
Yikes! I need an OS that's harder to get running than Linux like I need a hole in my head.
(A nicely written piece, I must say.)
Reply
Kristin Shoemaker said 2:23PM on 5-15-2008
Thanks! It came down to a "pick your battles" sort of decision. I knew it would bother me forever if I didn't figure out why it wasn't booting again on my computer of choice.
The networking... Since the drivers were recognized, I probably could have gotten it running eventually. But at that point... Eh, I guess the real mystery was solved and the magic blue smoke had escaped. And god, was I *tired.*
Matias Korhonen said 4:18PM on 5-15-2008
Just go with Gentoo...
Geir said 9:04AM on 5-16-2008
Gentoo doesn't have ZFS and DTrace.... :-)
Zach said 2:18PM on 5-15-2008
Oh my god I feel for you. That just sounded painful. Thank you for saving me from OpenSolaris.
Reply
Kristin Shoemaker said 2:25PM on 5-15-2008
I guess the bright side is I'm even more intimately acquainted with my BIOS than I was before?
It'll be interesting to see how/what it takes shape into. It seems a few people have everything work instantly... And other people have obstacles that don't particularly lend themselves to being overcome without a lot of tears.
bnitz said 11:21AM on 5-16-2008
Zach,
This is a good story, but there is something very suspicious about the conclusion. I wouldn't give up based on this review.
While there could be something broken with OpenSolaris 2008.05's USB device detection which hangs on some odd mouse during boot, why did it work the first time?
Given the symptoms:
-Live CD works the first time
-Live CD didn't work the second time
I'd save 5 hours and 55 minutes, ignore the typical wintel troubleshooting methodology (i.e. chaos), put your system back in its original state and burn a 2nd CD. I sure hope Kristin has tried this, but I didn't see anything in the article indicating this.
I'm not saying there aren't driver problems with some hardware (as there are on all Linux dists.) But if it worked the first time on Kristin's hardware, the chances of a stray scratch, smudge or dust particle screwed up one of the 56 million bits burned on the CD are much higher than the chances that the 1st Live CD messed up your PCs magic smoke. I will admit that this CAN happen, I once burned out a CDROM drive when a Microsoft Windows install went wrong.
Kristin Shoemaker said 11:35AM on 5-16-2008
Actually, I think I figured out the funny boot issue.
Before burning the disk again, I tried booting a few times on the P4 machine to my left. Booted perfectly, every one of the four or five times I tried. But not much of the hardware on it was supported out of the box, and the fact that it wasn't booting on the machine that booted once (and seemed to like most drivers better) was bothering me.
I didn't want to write about OpenSolaris on the P4, as it seemed a lot less likely that anything would run as expected. So I wanted to figure out why it wasn't booting on my machine.
Last night, thinking about it, I at least have a theory on it.
I had three USB devices plugged in at the first boot. One was superfluous to what I was doing (my mp3 player) and so I removed it before I booted the second time. That was the only thing that changed between boots. I am not even sure the disk was even removed from the drive on the (second) first boot failure.
Not sure how my motherboard is allocating USB assignments, or how OpenSolaris probes these things. The only thing I can possibly imagine is that the recognition of the mouse happened at a different point that first boot, avoiding the hang.
Ian said 2:51PM on 5-15-2008
Just try to change the hostname sometime. About 10 config files later, you'll know what I mean.
Reply
Geir said 6:00AM on 5-16-2008
Just run sys-unconfig.
jccalhoun said 2:58PM on 5-15-2008
to be fair the last time I reinstalled XP I had the exact same issues with my microsoft branded usb keyboard. It would start to install and hang. switch the usb legacy emulation still the same thing. dig out an old ps2 keyboard and it installs.
Reply
Matt Jacobsen said 4:22PM on 5-15-2008
I just went through a similar experience. I was hoping to use the one redeeming feature of Solaris, ZFS. But after all the setup woes, I discovered that due to a bug, ZFS cannot be used on external Firewire drives. Yeah, good to know. What's even better. This bug was reported over 2 years ago, and has languished because bug fixes need sponsors. So frustrating. It's like being trapped on Gilligan's Island and *really* needing to get off the island.
Reply
Geir said 9:04AM on 5-16-2008
You may want to switch to eSATA man....my external disks are much faster than my internal disks (I use 3 Gbps SATA disks).
Bufsabre said 7:09PM on 5-15-2008
the version of solaris ive been waiting for, for so long is nexenta, which is solaris with an ubuntu based os on top of it, great in theory but it is not the most stable thing ever and has very limited hardware support, sun has a long way with solaris to get it up to linux level, im ganna give this a try too but i picture more of the same
Reply
tbl said 2:40AM on 5-16-2008
I've got Opensolaris installed and running on 3 pc's in my homenetwork, including an amd64.
I've tried to install the new Ubuntu Hardy 8.04 on the amd64 on a partition, where i've used ubuntu for aproxiamately 2 years. But due to a bug in the kernel in 8.04 i cant install it. It doesn't recognize my sata disks any more.
https://bugs.launchpad.net/ubuntu/+source/linux/+bug/222176,
http://fosswire.com/2008/04/24/ubuntu-804-hardy-heron-review/
But Linux has many applications and benefits, so I have to use it sometimes. So how did i install the new Ubuntu? Virtualbox.
NetBSD is also a very nice system, with a good package system.
Reply
Geir said 5:58AM on 5-16-2008
Did you try the Linux boot option all-generic-ide?
Geir said 6:24AM on 5-16-2008
I'm surprised to see how negative this article was on behalf of OpenSolaris, I think most operating systems will have issues on some systems. To hang out OpenSolaris due to failure on one machine like here do not seem fair.
Anyway, I'm not here to defend OpenSolaris, even though it has some heck of a nice features like ZFS, DTrace, and the TCP/IP stack. What you should have done is to test it on more than one system, and you should also use the [OpenSolaris] community to request help and report the issues you find --not just blow off steam and debunk the OS. I don't think it would take 21,600 seconds before you got help if you requested it...
On the other hand, the article is well written and entertaining. Just hope it won't scare off others from trying it.
Reply
Kristin Shoemaker said 8:20AM on 5-16-2008
Thanks, Geir!
I did actually try it on every machine I had available to me (working from home limits my options). An older P4 machine booted consistently, and I was more than willing to just base the review on that... Except the hardware on that machine (network, most notably) was not even as well supported as the machine I eventually used. The ethernet driver for the network card on the P4 machine had a long discussion about installation and configuration, and out of six people or so in the thread, it seemed one got it to work reliably. I considered trying it on my Eee, as I have heard the wired ethernet works well.
Much of those six hours were spent Googling for help, alas, because I really desperately wanted to give this LiveCD a healthy chance. I also know well that most people are much less patient (and thereby, much more sane) and if they couldn't get something to work the way it needed to after a much less significant amount of time, they'd have cursed wasting a disk and tossed the CD-R in the trash.
It didn't seem like something I needed to ask on support forums. People had asked very similar questions. Many, very many, were unresolved (though there were some nice tips on how to further troubleshoot, and that helped greatly.)
I've had my fair share of other OSs not working as planned on hardware. It does happen. An early Fedora release would not boot come hell nor high water on my former desktop. I've had problems with Windows.
Speaking reasonably, most people just looking to try out an alternative desktop don't want to search for answers for as long as I did... Did everything in my power to make this work, without looking too neurotic about my compulsive want to get things to fly right.
It's why I recommended steering clear of *this* release. Will the next one/the one after that be as difficult? Most likely not. But I'd still have to say, as it stands, it's not a bad release to forgo in favor of some other open UNIXes.
Geir said 9:03AM on 5-16-2008
Yeah, I know how you feel. I had a big issue on the "regular" Solaris on my notebook before too. It is a pain when things don't work, but I've never spent 6 hours trying to resolve something. I either try all reasonable options and then either debug it myself (if it is open source) or ask or check if someone in the community has encountered the issue before. I have to say I have had more issues with RHEL/CentOS detection of hardware though, but usually there are ways around every issue unless it is something that is flat out not supported.
Anyway, OpenSolaris is to Solaris what Fedora is to RHEL/CentOS, so when you go for something cutting edge then be prepared to bleed... :-) Wow, that sounded very cheesy....but I think you know what I mean.
You did manage to highlight the weakest side of OpenSolaris though, namely the comparatively modest amount of supported hardware.
Thanks for your clarification.
Kristin Shoemaker said 9:07AM on 5-16-2008
Yeah, it was a bummer, because I really did want this to work. I don't have enough programming knowledge to make major alterations to code... Certainly the -k and -v switches in GRUB made life a lot easier when it came down to figuring out what is making this so touchy...
I would have asked a question or two... but whereas it seemed very similar problems had come up, and were resolved with varying amounts of success, it bothered me to ask the questions again.
I did see advice on the forums that I tried... Memory timing, disabling PnP, or, yes, even removing RAM (okay, I didn't try removing RAM. I'll mess with my BIOS. But it's too much effort to find the screwdriver, crack the case, and pop out the memory module.)
There seemed to be someone on the forum who did have a hang up on boot, revealed by the -v switch, but it was an entirely different piece of hardware causing the issue. Don't know if he actually was able to fix it... I was lucky, actually, it was just a mouse issue and my house is filled with adapters. :)