I have read in the RAID mailing list archives that it is possible to take an existing ext2 fs and an empty disk and make a RAID 1 array containing the data on the existing fs.

The way they advise doing this is in the 4th section of this howto:
http://www.linuxdoc.org/HOWTO/Boot+Root+Raid+LILO.html

They advise making a raidtab with the new empty disk as a disk on the array and the existing partition listed as a failed drive.
They then say to initialize the array as usual, start it up, copy all the files from the existing partition to it, then stop the raid. You then edit the raidtab and list the failed disk as a good disk and turn the array back on.

This seems to me like you are making a single drive raid partition, copying all the crap to it, then adding the original disk to the raid array and the old data gets overwritten with the copy from the new drive.

This is great except I am extremely anxious to perform a format with my existing partition listed at all!! Even as a failed drive!! Is that really necessary to list it as a failed drive? Couldn't I simply not list it at all, then add it after the file copy?

I have one other question... if a software RAID 1 array fails, or one disk fails, can I mount one of the RAID partitions as a standard ext2 fs? I am kind of anxious on the issue of what should happen if the machine crashes or one disk fails. It would be extremely reassuring if I could mount either mirror as a standard fs.

Also, is are all ext3 journalling options supported over software RAID 1?

I believe I may have found my own answer. The drive needs to be listed as a failed drive so that the superblock has that drive info when the new drive is initialized into the RAID array. This is actually hinted at in the man page for raidtab, so I'm more comfortable with the idea now. Am I correct?

chunk size would only be valid or important if you were doing RAID 0 or some other striping RAID.. It would not affect anything with a RAID 1 setup, correct?