Solaris 10 fsck

Today a POS E250 that runs some Sunray kiosks crapped out for no particular reason. I am working from home today, so I tried to SSH into the machine, with no luck. I next tried telnet, also no luck. Ping worked though.

Uh oh...hung system.

I had to get a Windows admin on-site to do a cold boot on the machine, which did not help. I was all set to drive into the office on my flex-day...again...and then I remembered the magic of RSC. E250's come with an RSC (Remote System Control) card and, further, have a neato little utility that shows you a GUI version of the front panel and also allows a command console, even when the server is screwed up or off. Newer versions of RSC are web-based.

Anyway, as I suspected, the idiot server was hung waiting for an fsck to be run on a particular slice. How to handle this? Well, first you will need to enter the root password to get into single-user mode. Then you will need to check the appropriate slice. In this case, the slice having difficulty was listed as:

/dev/dsk/c1t8d0s6

So, now we run fsck on that slice. There are two important points here. First, fsck requires the raw device (if the slice is larger than 2Gb), which means you would run it on:

/dev/rdsk/c1t8d0s6

The second point is the -y flag. If you do not use the -y flag, you will spend the rest of your life pressing 'y' to fix busted inodes. So, the command you want is:

fsck -F ufs -y /dev/rdsk/c1t8d0s6

or

fsck -F ufs -o f,p /dev/rdsk/c1t8d0s6

Both will have the same result. The second method may be slightly safer because the options are based upon the filesystem. In this case, f=force and p=preen (non-interactive mode)

The '-F' is optional and tells what type of filesystem is in use. The default filesystem for the machine is defined in /etc/default/fs and fsck will use that value if you do not provide one. I use the -F out of habit and to be sure I get it right. YMMV.

Once the fsck completes you can either reboot the machine or type exit to continue the boot process. This is especially useful if you have more than one filesystem that is botched, or if you are not sure of the status of the remaining filesystems. I usually use the exit feature to make sure I got everything and then do a graceful reboot to be sure.

-TheDave

Swap Space

So, today I am building a machine to use as a test box for ensuring that our various databases can survive application upgrades, database upgrades, and OS patches that have been accumulating for the last couple of months. A fairly routine install of Solaris 10 on a 280R went well until it came to the swap space.

I never allow Solaris to automatically lay out my filesystems as having a 600MB / slice and a 70GB /export slice is entirely useless in my environment. So, I went ahead and chopped it up as I saw fit and let the installer work its magic.

The came the time to set up the swap slice. I generally install the swap area at slice 0 and use the beginning cylinders of the disk for the space for the sake of speed. Especially on a database machine, that can be important. So, the first thing I do is go into format and add my swap slice. Then I attempt to add the newly made slice to my running swap config:

# swap -a /dev/rdsk/c0t0d0s0
/dev/rdsk/c0t0d0s0: No such file or directory

oops...I should not be using the raw device. I try again, this time using the actual device path:

# swap -a /dev/dsk/c0t0d0s0
/dev/dsk/c0t0d0s0: No such file or directory

What the hell? OK, time for a sanity check:

# man swap
Reformatting page. Please wait.... done

Before reading a word, I realize that I failed to make a filesystem on the newly created slice.

# newfs -v /dev/rdsk/c0t0d0s0
newfs: /dev/rdsk/c0t0d0s0: No such file or directory

Well, this is NO good. Maybe I forgot to write the label when I formatted the disk? A simple check with format shows that the slice is, indeed, still there. Did I use the raw device when attempting the newfs? Let's try again, just to be sure:

# newfs -v /dev/rdsk/c0t0d0s0
newfs: /dev/rdsk/c0t0d0s0: No such file or directory

Damn. What the hell do I do now?

To make it short, I spend the next 45 minutes reading various man pages and web pages to try to figure out what my problem is. I even had an abortive experiment with mkfs, which is not useful in this situation. Then during my umpteenth view of the output from format I see a small detail that I missed. See if you can catch it:

# format
Searching for disks...done

AVAILABLE DISK SELECTIONS:
0. c1t0d0
1. c1t1d0

Specify disk (enter its number):

Give up?

I tried:

# swap -a /dev/dsk/c0t0d0s0
/dev/dsk/c0t0d0s0: No such file or directory

I was using the wrong damn controller number. The lesson here, ladies and gentlemen, is to PAY ATTENTION to what you are doing so that a 1 minute effort doesn't end up taking a whole hour of your life from you.

Incidentally, if you wanted to make the new swap space permanent across reboots, you would need to add the following line to /etc/vfstab:

/dev/dsk/c1t0d0s0 - - swap - no -

You do not NEED to reboot at this point, but this is a test machine and a new build so I did so anyway.

-TheDave