Solaris on x86 - Problem Booting Solaris 10/x86 After Lockup
I have a Sun Ultra 20 with Solaris 10/x86 running at home. I recently reconfigured my DSL modem/firewall and rebooted the modem/firewall. During this, my Ultra 20 locked up. I turned the power off on the Ultra 20, then on again and proceeded through the normal boot sequence. The Ultra 20 starts the boot sequence from the hard drive. However, after the hostname displays during bootup and before the login screen displays, I get this error:
svc.startd[7] svc:/system/power:default: Method "/lib/svc/method/svc-power start" Failed with exit status 96
There are a bunch of other errors which display (all svc related) before the login screen displays. When the login screen tries to display, all I see is a black screen with an hourglass for a mouse pointer.
I've tried to (hurry up and) login to the console before the login screen displays. Most of the time when I do this, the black screen displays briefly, then I go back to text mode...but I don't have a prompt and cannot execute any commands, I only get more svc errors. Once (and only once) I got to a Solaris command prompt through this process. I ran sync (as I've had previous problems booting after a power outage or reconfiguration of my firewall and subsequent Solaris lockup) and a sync solves my problem. This time, after the sync I ran reboot, however this time it didn't fix my problem.
Telnet is disabled (per security announcements) on the Ultra 20 and I cannot ssh to the machine. I believe the ssh daemon is not running yet. I can ping the machine and my web server runs as I can bring up a web page from another machine. Stop+A does not work on x86 machines.
Some of the errors (and information I've found on the Internet) state a 'svcs -x' should help me diagnose the problem, but at this point I cannot get to a prompt to do anything.
I do not believe this machine came with a copy of Solaris 10 on CD. I have a 'Sun Ultra 20 Workstation Supplemental CD v1.0' which contians some DOS utilities, but nothing which can help me with this problem.
Any help is appreciated.
[2100 byte] By [
darryl_ma] at [2007-11-26 23:09:09]
# 2
Alan,
thanks for the suggestion. I ran 'b -s' when booting. I saw this, which was different:
Booting to milestone "milestone/single-user:default".
I saw this:
Root password for system maintenance (control-d to bypass):
I entered the root password, which was accepted, then everything was the same as before. I received a bunch of svc errors, then the black screen and hourglass. If I hurried and tried to login before the black screen, I was kicked back to text mode, then nothing.
Any other suggestions?
# 5
Alan,
can I fsck without logging in? My main problem is that I cannot even login, so I have no chance of fixing the problem.
BTW, I'm in the process of burning CD's right now to see if I can boot via a CD, mount the drive, and fix it that way. Not what I really want to do, but I see no other option.
-Darryl
# 7
Alan,
yes...I was hoping to be able to find a way to login without having to boot from a CD, but alas... :(
An update...I burned the 6 Solaris 10/x86 CD's (not sure which ones I'd need). I booted from disk 1 and ran fsck on the HD...of course, errors were found. I corrected them. I then tried booting from the HD again. I was told there were errors in the repository database (/etc/svc/repository.db) and to run /lib/svc/bin/restore_repository. I was then asked for the root password. I entered it, but received the same exact message again (re: repository corruption). It turns out that root's home directory was removed. So I had to reboot using the CD again, mount the drive, create a home directory for root, then reboot to disk again. After receiving the repository corruption message and root password request, I was able to login...finally. I then ran restore_repository, picked the earliest boot version I could find, and was able to boot and login normally. Phew.
What scares me the most is that all of the home directories I had in /export/home (there were several) are completely gone. Fortunately, I really had no data there...certainly no important data. I was planning on moving the directory of the login I use most from its current location of /local/home to /export/home (so autofs picks it up and I can access it from /home). But now I'm hesitant to as everything in /export/home got wiped out. I have quite a bit of important data in the home directory of my login. Possibly autofs caused the problem...not sure.
Thanks for your help.
-Darryl
# 8
was able to boot and login normally. Phew.
everything in /export/home got wiped out. I have
quite a bit of important data in the home directory
of my login. Possibly autofs caused the
problem...not sure.
They gots this thing nowadays that be calling backups. They supposed to hep in situations like this. :-)
Anyways, you might want to run your files against the Solaris Fingerprint database to see if there's any funny business going on that you don't know about.
Not sure what happened to the file system but at least you're up and running.
alan
# 9
Heh heh heh...yes, I have backups of all my data. While it's important, it is also replaceable. What isn't is the time and effort I've put into downloading, building, installing, and setting up various services I'm using on this machine. I'd rather not have to go through this again.
I'll look into the SFD. This looks like a good service. Thanks for the tip.
-Darryl
