I have set up a RAID 1 software raid system.
What I'd like to be able to do is when performing upgrades etc I can split the mirror off and keep one mirror as a backup.
If the upgrade works I then add in the old disk again.
If it fails I'd like to be able to rebuild the array using the OLD mirror that I kept.

I cant find any info on this. I know I can fail the drive with mdadm, but my understanding is that the data on this this drive can't then be used again. Is that correct or can the bad upgrade disk be failed and then the old disk un-failed?

Thanks

Still on the zero reply threads list. hmm. I want to jump in and give you the steps, but my linux mdadm experience is scant and mostly second hand. I've done a lot of raid, mirroring, breaking, rebuilding and upgrading, but all in Solaris (just did it today). I can say that what you are suggesting is common practice. We had a tutorial on it in our weekly sysadmin meeting a couple of months ago (with linux and mdadm).

I could check the man pages and google it, but you could do that too. Google mdadm, with split or break, and upgrade or backup. That gives you a couple of combinations, and it turns up a lot of threads that are close enough to be useful. Or toss howto into the search. "howto mdadm" turns up some useful material.

Or maybe someone will pop up and give the steps here. Should only be a half dozen or so steps for the whole process.

That seems totally possible with mdadmn. I haven't done that since we usually have "staging" servers for that type of purpose. Having an extra computer/server for testing reduces human error and saves a lot of time. It's also cleaner. However, I can understand if the budget is not there for this exercise.

You can't force a rebuild against the disk that has the updates you want? What does the error say when you try to do a force rebuild?

-twantrd

Thanks Guys. I honestly have googled it but unfortunately the words split and break all have different meanings so bring up lots of articles such as 'disk breaks' and 'split into 3 disks'
I wasn't even sure this was possible, hence my question.
I was also hoping someone on here would do it reguilarly and be fully versed with how to do so.
Now that I know it is possible I will keep googling, and playing about.

I haven't tried forcing a rebuild against the good disk yet. I'm very new to all this so please bear with me.

So I have a good backup which I have 'failed' manually. This actually stays intact then does it?

What would the next step be - fail the other disk? Drop the array completely? Then how do I create a new array using the old disk without creating it as a fresh blank array?

Thanks again. As I said I'm new to this and still trying to work out exactly how it all works so any pointers you can give me would be greatly appreciated.

Hi,

Did u find out how to do this?

I am facing the same situation. Need to distupgrade from debian sarge a software raid 1 system.

Situation: Raid 1, mirrored system running on 2 identical disks.
Steps: (as I have figured)
-I power down
-I unplug one disk
-I power up
-I do the dist-upgrade
-I power down
-I insert previously unplugged disk

Now comes the problem:
-If the upgrade was successful I obviously want the upgraded disk to be synced to the previously unplugged disk

-On the other hand if the upgrade failed I want the unplugged disk to be synced to the upgraded disk.

How does the software raid determine which disks data to use in the array and which disk to sync to which? Especially if both have working copies of the data but different versions.

Thankful for any help.

Probably figured it out.
Seems like I have to wipe the partition table of the drive I dont want to be used?