Solaris 10 quality
Quite a few people who have Solaris 10 auto patched found their system broken. If there was such a problem with Microsoft Windows, you would have seen the news in CNN Headline News and People's Daily (of PRC).
But this is Solaris 10, and INQ did not bother to report it.
My Solaris 10 installation is having an error during boot, asking me to enter root password to enter maintenance mode -- this is on a virtual machine...Then what? I ran fsck and stuff like that, the problem persists. Running "svcs -s" showing a long list of services not started... I don't have the time to investigate further. I see others having similar problems (system in continuous rebooting loop), here too (system hangs on boot). And more crap out here(SUN hardware, system hangs after reboot), more failure here(HP hardware), more failure here (DELL hardware), failure to install (HP notebook), SATA drive support issue, system hangs (no final resolution).... SUN brags that Solaris 10 embraces HP, DELL and IBM customers, I seriously doubt it, as I found Solaris 10's hardware support lacking.
I never had such problems with Linux. I don't remmeber when was the last time I had to reinstall the Linux kernel after a power failure. With ext3, Linux can boot without root intervention even when filesystem inconsistency occurs. Performance-wise, Linux is simply unbeatable.
I don't see a C/C++ compiler with the default Solaris 10 installation. Javac is there though.
posted by Sharikou, Ph. D. @ 12/22/2006 10:58:00 AM
28 Comments:
All I learnt from this "article" is that you have no ideas about how to administer Solaris.
Also, have you missed all those Ubuntu updates that screwed up X configs or drivers?
All I learnt from this "article" is that you have no ideas about how to administer Solaris.
Looks like every solaris 10 installation needs a $100K/year sysadmin to keep it running.
jj_japan
Posts:2
Registered: 12/11/06 Re: solaris 10 can't boot up after system reboot
Dec 11, 2006 9:36 PM (reply 2 of 3)
Hello!
This is not a reply to offer a solution. We just have a similar problem on three of our SunFireX2100 with pre-installed Solaris 10. After configuring Solaris 10 in one machine, its works out fine. We use the following commands to enable the graphics mode:
#eeprom console=graphics
#/usr/dt/bin/dtlogin -e
#reboot
At this point, it works out fine, but after rebooting the system again, the system hangs-up. We tried to start up using Solaris Failsafe mode and tried to reboot again selecting the installed OS (Solaris 10 1/06 s10x_u1wos_19a_X86). But still the systems hangs-up at the Solaris Welcome Screen.
We resorted to use sys-unconfig but to no avail. When we run sys-unconfig it returned an error in unconfig_files stating it cannot find or delete /etc/resolv.conf.
The other two machines have the same situation now after second reboot.
What could be the problem? We would appreciate it very much if someone can offer solution.
Thanks in advance!
~~jj~~
you should know that the "problems" you described mostly have nothing to do with Solaris "quality".
User friendliness, perhaps.
you should know that the "problems" you described mostly have nothing to do with Solaris "quality".
No. When your enterprise OS hangs or enters an infinite loop of rebooting. It's crap. Period. SUN claims that Solaris 10 can help HP or DELL users, did they even buy a HP or DELL box to see if Solaris boots?
I don't understand... Sharikou why you attempt to do down about Solaris? Solaris is a Sun product, Sun is actually a big supporter of AMD Opteron technology you love so much. I don't understand, perhaps is this an attempt to drive this blog on the right side of the really "Pervasive 64 bit computing" instead to say only "Intel sucks"? xcuse me, I know, I'm a bad English writer.
In HP webpage, for each machine, you can see which OSs the machine is certified to work wth. I can bet no HP laptop is certified to run Solaris 10.
When you buy Solaris 10 through HP (or Dell, or whichever) to run in a machine certified for it, You will also get Customer support, and the issue will be resolved real fast.
Solaris 10 never was meant to be a consumer OS, but rather a server OS, and therefore, it will requiere a competent sysadmin.
The only point where I may agree with you, is that Solaris 10 sould not have problems in well knonwn Virtual Machines like Xen, Virtual PC, VMWare, etc (not some obscure FOSS project). But, since you are not a Solaris Sysadmin, it may as well be a problem with the original install.
Slow news day at Ph. D. Central, I guess... ;-)
Salud!
No such problem here. Solaris 10 update 2 runs great. Yet to try out update 3. Perhaps you need to post your problem in some Solaris help alias rather than whining in your blog. How about comp.unix.solaris? Unless you just want to diss Solaris, of course!
osgeek
don't understand... Sharikou why you attempt to do down about Solaris? Solaris is a Sun product, Sun is actually a big supporter of AMD Opteron technology you love so much.
I am only stating the truth. SUN should thank me for telling them what's wrong and what to fix, instead of dancing on a pitiful meagre 80K user base. SUN's only chance for survival is Solaris. Right now, Solaris 10 simply sucks. In a previous post, I pointed out the situation with usability. Here, I point out the obvious stability and quality issues of the OS kernel. No admin can fix problems of a hanging system and continuous rebooting system -- short of reinstalling the OS. SUN had a three-year window of oppurtunity to get into the AMD64 market. It was simply not quick enough. The galaxy servers took ages to show up. And Solaris 10 won't boot on Hp Dl385 and BL25p...
But, since you are not a Solaris Sysadmin, it may as well be a problem with the original install.
So, you are saying that I, a person with an advanced degree in computer science, who used Solaris for years, who knows about operating systems inside and out, can't manage Solaris 10 myself. If you are telling me that I have to read 500 pages of SUN manual to learn all the crappy command line options to get the sucker to boot, I tell you most people don't have the patience. I at least have some interest on Solaris, most other people simply won't give a damn and choose something else, such as Linux. Some say Solaris 10 does not run on a HP notebook (or DL385, BL25p), guess what? Linux runs perfectly fine on notebooks, a Turion 64 X2 is more powerful than an UltraSparc no matter how you measure it.
Wow, we should thank you for such tips? Wow ... what should we do than with the people that really contribute code ? With people that files bug with it? Weight their weight in gold? You have obviously not the problem of a small ego.
And to your list: Should i really start to list problem with Windows. Should i really start to list problems with Linux. Should i really start to comment the problem of empty support intersections between a linux distribution, the used hardware and the application and an driver? Should i really talk about a company, that put it´s file on one of the glorified file systems of linux and was 20 days later in the lucky position to have a tested backup and restore procedure? Do yourself a favour and look into the Hardware compatibility list. And stop about talking of solaris. There is a proverb in germany "Wenn man keine Ahnung hat, einfach mal Fresse halten" or in english "Without a clue shut up". And maybe you have some clue about opterons. Operating systems, their usages and their markets are definitly not your field of expertise.
"When your enterprise OS hangs or enters an infinite loop of rebooting. It's crap."
Oh, certainly, as if a redhat or suse would never enter an infinite loop of rebooting? A Windows update would sometimes get my box into such state, too.
Had you been to a BSD forum you'd also find tens of problems with each new release, yet BSD (IMO, too) is one of the most solid OS out there (BTW FreeBSD AFAIK doesn't even have automatic update; yet you're simply ignorant to say FreeBSD is 'crappy').
As I said, you're simply confused of user friendliness with OS quality.
"SUN should thank me for telling them what's wrong and what to fix, instead of dancing on a pitiful meagre 80K user base"
sun dont have 80K user base. someone told you already, I can repeat if PhD in advanced computer science is not enough to get it right from first time - noone in his mind in corporate world runs auto update. so number of auto updating users is number of lamers who downloaded OS for fun. thus 80k (or rather 143k) is not that small number for quite special corpotate OS.
5 failed to boot reports from google is nothing, even for 80k user base. its easy to find same booting issues for any kind of os and hardware combination. hardware can be broken, solaris is very demanding towards hardware, ppl not very professional in average.
and finally. apple will use 2 core sol10 technologies in nex gen macosx, leopard. zfs and dtrace. in case u missed that.
And to your list: Should i really start to list problem with Windows. Should i really start to list problems with Linux. Should i really start to comment the problem of empty support intersections between a linux distribution, the used hardware and the application and an driver?
I assume you are from SUN, and your atitude is a bad sign. Linux has problems for sure, Windows have more problems. Linux people listen and they make quick changes. Microsoft doesn't bother, because people simply love Windows--no one else created a mildly competitive product. People voted for Windows with their money. People voted Linux with their installations. Why? At least, both OSes can be installed on almost any x86 hardware without a glitch and run merrily. You can't say that about Solaris 10. The links to problems people experienced are not random hardware errors, but software defects.
I saw one poster saying Apple will base its OS on Solaris 10, if that's true, why do we need SUN? I suggested that Linux and Windows should take the best of Solaris 10 and incorporate those features--SUN spent billions on Solaris 10 development, it would be a shame if such money is wasted. But, frankly, I wouldn't run my servers on solaris 10 as it is today. I have run Linux for years, never had any major problems such as unable to boot. Some talked about FreeBSD's problems, but FreeBSD is a marginalized project, Linux user base and developer base is 100x of BSD (BSD install base is 50x of Solaris 10). Look at Linux today, with every update, I see continuous improvement. Linux is mankind's only hope to stop Microsoft. SUN should join Linux for real, instead of fighting it. Why can't SUN dedicate 100 programmers to port ZFS and DTrace to Linux? Instead, ZFS is written with all those Solaris 10 specific stuff...
I always praised Solaris 10's features, it's a nice technology. SUN engineers are among the brightest. But that's not good enough. SUN needs shrewd leadership in executing its business. Empty talk and endless bragging are meaningless. If Solaris 10 is good, make it work!
sun dont have 80K user base. someone told you already, I can repeat if PhD in advanced computer science is not enough to get it right from first time - noone in his mind in corporate world runs auto update.
80K is the number SUN exec were dancing for. There is no other number (oh, I did mention the 5 million downloads to which I contributed at least 5). So 80K is the reference, and it's a pitiful negligible immaterial number--especially considering the hype and length of time since Solaris 10 is out.
The idea of solaris 10 is allow anyone who has an interest to download it for free, and hopefully they will keep it. You can see it's not happening. Even if I 5x the 80K number, it's still insignificant. A startup does better than that with a brand new OS. You have to remember SUN has only 10% of the server market in terms of revenue, in terms of units, SUN has less than 2% of the market. There are 10s of millions of servers running.
Sharikou,
Can you comment on my predictions for 2007?
itkitty.blogspot.com
"So, you are saying that I, a person with an advanced degree in computer science, who used Solaris for years, who knows about operating systems inside and out, can't manage Solaris 10 myself."
Have you ever thought that your degree isn't really worth its name? Judging from the things you've written here I would certainly think so.
SUN should join Linux for real, instead of fighting it. Why can't SUN dedicate 100 programmers to port ZFS and DTrace to Linux? Instead, ZFS is written with all those Solaris 10 specific stuff...
Get lost Sharikou. Stick to Intel fake maximum power consumption and the Intel uber cache processor. Have you any idea what the current status of XFS is in the Linux kernel? The code is getting worse and worse because of Linux's architecture for which work arounds had to be made. Very core and senior XFS developers have left the Linux porting project thanks also to the stubborness of the core Linux development community. Witness the difficulty getting advanced filesystem features into the linux kernel that Hans Reiser has faced and continues to face. He and his team had to strip off almost all the good stuff just to get reiser4 into the Andrew Morton kernel.
Linux is a piece of crap compared to the Solaris kernel. I, for one, welcome the GPL'ing of Solaris and will happily support Solaris in favour of Linux boxen now that Solaris is available on commodity PCs. You just have to pick the right hardware like AMD Opterons on Tyan motherboards that use Nvidia chipsets or AMD chipsets in the computers you build and stuff into data centres for the moment. (Read: NO INTEL XEONS THAT ROAST HARD DRIVES) Linux does not match the Solaris in kernel reporting facilities nor virtual memory management nor I/O. Porting ZFS or Dtrace will be a major pain just like the SGI guys found out trying to port XFS from Irix to Linux.
Advanced degree in computer science? I almost fell out of my chair laughing. Why anyone still buys this line of crap is utterly beyond me.
He is NOT a ph. d.
He is NOT a journalist.
He IS a disgrunteld ex-Intel employee who was fired for systems abuse. This has been proven and established on many other sites. Search for it.
Linux is a piece of crap compared to the Solaris kernel.
Linux code is of higher quality, tens of thousands of developers have examined Linux code and that made it quite solid. Solaris 10 has nice ideas but code quality is obviously lacking as demonstrated by the links in this article. With IBM and others spending millions on Linux, it will get better. SUN should really give up the fight against Linux, instead it should contribute to Linux, otherwise Solaris 10 will be just technology wasted.
We see all the hype about Solaris 10, but those are just pure theoretical arguments. There is no research showing Solaris 10 more reliable or better performing than Linux.
I think you're on to something, regarding the lack of a C/C++ compiler and the presence of the Java compiler.
Java's promise is "write once, run everywhere." The advantages it provides to a software company:
1) you never have to let your source code outside of your doors, so no one else will ever see how crappy your code is.
2) the code runs inside a virtual machine, so you never have to worry about crashing anything with your poorly written code.
Today the advantages come at the cost of a very slow JVM that has to be loaded with a lot of smarts (JIT compiler). These costs are a huge advantage for a hardware company, because they make your customers want to buy faster hardware to offset the slow JVM.
The concept is interesting, but in practical terms it's not very different from sending out your programs in source form, requiring a standard compiler and set of libraries to exist on the target machine, and compiling the code on the spot. The main difference is that the latter approach would be more efficient, and the resulting programs would all execute a lot faster. Also, obviously the latter approach puts more power in the hands of the customer.
Too bad. Even if Java flourishes as an open source project, there's no getting around the fact that running inside a JVM costs too much. The only way to get rid of those costs is to run native code. The fact that Solaris 10 has containers and zones tells me that Sun has realized their JVM direction is a dead-end. That's probably the real reason they've finally decided to loosen their grip on Java. Virtual machines that run at 10x slower (at best) just aren't interesting; virtual machines that run at nearly 100% native speed are the only way to go.
Linux code is of higher quality, tens of thousands of developers have examined Linux code and that made it quite solid
ROTFL. In the face of calls for less new feature coding and more bug fixing from Linux developers like Andrew Morton I declare your statement complete nonsense. Oh, there may be certain parts of Linux that are good and solid but I am afraid those parts have not been touched for years and the kernel's architecture is wanting because you will only find good solid code on versions lower than 2.6 and even then they still have issues which is why they got changes in 2.6. Made it quite solid he says. Funny that a notable few on LKML feel otherwise.
There is no research showing Solaris 10 more reliable or better performing than Linux.
Not research, just experience. Solaris proves itself both performance wise and reliability wise. Look at this comparison of Asterisk under Centos 3.7 and Solaris 10 on the same hardware. Note that Asterisk was first written for Linux and is still primarily developed on and for a Linux system
http://www.thrallingpenguin.com/articles/asterisk-solaris.htm
Solaris is carrier-grade and Linux is not. Period.
http://www.thrallingpenguin.com/articles/asterisk-solaris.htm
Solaris is carrier-grade and Linux is not. Period.
The test is a retarded. The tester is a retard. The machine (a celeron) had only 512MB RAM. As we all know, CentOS's default installation enables a lot of services which takes up a lot of RAM, Solaris 10 is barebones OS with little extra to install. The so called "fault" in CentOS was apparently "out of memory" error.
"80K is the number SUN exec were dancing for. There is no other number (oh, I did mention the 5 million downloads to which I contributed at least 5)."
corporate os is not downloaded and installed. in fact all corporate users gets new os version with new hardware. so if u want real number of production sol10 installations, get volumes of amd and ultraT1 sales from sun, plus there some arbitrary percent of ultra sparc sales. that would be number of real sol10 users. and trust me, noone has autoupdate turned on. most of them dont have even internet access.
"The idea of solaris 10 is allow anyone who has an interest to download it for free, and hopefully they will keep it."
sol was available for free long time ago before 10.
at least for sure starting from 2.6, but still, free version was always adressed to universities and alikes, not real customers.
"A startup does better than that with a brand new OS."
depends how do u measure it. if u measure corporate OSes, add here HPUX, Irix, etc, solaris numbers are top ones. if you care about consumer OSes, then it obviously lacks. besides ugly installer solaris dont have built-in DVD movie creator and so on, thats even more serious problem with it.
"You have to remember SUN has only 10% of the server market in terms of revenue, in terms of units, SUN has less than 2% of the market."
SUN is not in its best shape now. however having 2% in volumes, they have 10% of revenue, its dream of any CEO/CFO, no? high marhin products?
u suggest to discontinue sol 10 and port all advanced technologies to linux. and what will sun do after that? compete with DELL in lowest server margin?
depends how do u measure it. if u measure corporate OSes, add here HPUX, Irix, etc, solaris numbers are top ones. if you care about consumer OSes, then it obviously lacks.
You ddidn't mention the two top enterprise servers, Linux--backed by IBM, HP and DELL. Windows Enterprise--backed by IBM, HP and DELL.
Both are 50x of Solaris 10 in terms of install base in enterprise computing env. The people who are running Solaris 10 are mostly those who stuck with the SPARC. Ask SUN how much % of their x86 server customers are using Solaris 10? I bet it's a small number. Even SUN the top customer, the Tokyo inst of tech who bought a number of x4600 and x4500 to make a supercomputer does not use Solaris. In fact, they use Linux even on x4500.
Solaris was fading and Solaris 10 lacks the quality to be truly attractive.
Don't be mistaken EXT3 will still requires manual "root" intervention when select corruption occurs.
The test is a retarded. The tester is a retard. The machine (a celeron) had only 512MB RAM. As we all know, CentOS's default installation enables a lot of services which takes up a lot of RAM, Solaris 10 is barebones OS with little extra to install. The so called "fault" in CentOS was apparently "out of memory" error.
From the article: During the high loads placed by SIPp, Asterisk segmentation faulted under CentOS but never crashed on the Solaris operating system.
As we all know (you do don't you?), Linux does a great job of shoving inactive processes into swap so if the tester really is retarded and did not shut down unnecessary services it would still mean squat. For your information, a segmentation fault is not the same as an OOM error. A segmentation fault occurs due to faulty RAM (not the case since Solaris 10 used the same box) or due to bugs in either the code and or the libraries it loads. Okay, strictly speaking this probably means that the Linux kernel is not at fault here but when you speak of Linux, we refer not just to the kernel but the whole operating system.
GNU/Linux is nowhere near Solaris' stability. Period.
I don't see a C/C++ compiler with the default Solaris 10 installation. Javac is there though.
You can get Sun Studio for free like Solaris 10. Just like you don't get gcc installed by default on most Linux distros.
Got Nexenta installed and running on a Dell GX280. Solaris runs on Dells
Post a Comment
<< Home