Consequences of tes...
 
Notifications
Clear all

Consequences of testing mirror failure

6 Posts
3 Users
0 Likes
5,864 Views
(@notio)
Posts: 12
Member
Topic starter
 

Having just created a 5TB mirror volume, I decided to pull the secondary drive and test failure...

* When I connected the pulled secondary drive to the same computer using a different enclosure, SR knew it was a secondary mirror and did the expected thing of mounting the drive and verifying the mirror. This took less than two minutes. Most excellent.

* I then unmounted the mirror volume ("MyMedia") and moved the (removed) secondary mirror to another computer, and mounted the removed secondary via the Finder. Happily, all the files were there, and I could use the drive normally (though I didn't do anything other than browse the file system). This was the test result I wanted to see, so again, most excellent.

* I then reinstalled the secondary mirror in the original enclosure (note that the mirror volume was unmounted at this point) and this is when I became confused. See screenshot for the current status.

- SR mounted two volumes, both named MyMedia, both missing the secondary disk of the mirror pair. I.e. it appears SR thinks there are two primary mirror volumes, both named MyMedia, both missing their secondary.

- I don't see a way to tell SR "Hey, this volume is actually the secondary of this other volume."

What it looks like I *might* want to do is "Remove missing secondary disk" -- then add this same disk back as the secondary. Would this have the expected outcome?

Obviously I'd like to avoid re-copying the 5TB over to start over, but if that's required it's okay, as the only way this would happen IRL is under a failure condition much worse than what I've simulated.

Advise appreciated!

 
Posted : 31/12/2018 12:38 pm
(@softraid-support)
Posts: 8006
Member Admin
 

What you did was fine, but if you want to move a disk to another computer, move the primary. this is because when you connected the secondary, it failed over. Then SoftRAID cannot know which disk should be the "real" primary.

In your current case, delete the duplicate volume on the ex secondary. then "remove missing secondary disks" and "Add secondary disk" the volume will rebuild.

 
Posted : 01/01/2019 2:45 pm
(@notio)
Posts: 12
Member
Topic starter
 

Ok, this worked great, thanks for your help. (Appreciate the tip on pulling the primary.)

What you did was fine, but if you want to move a disk to another computer, move the primary. this is because when you connected the secondary, it failed over. Then SoftRAID cannot know which disk should be the "real" primary.

In your current case, delete the duplicate volume on the ex secondary. then "remove missing secondary disks" and "Add secondary disk" the volume will rebuild.

 
Posted : 02/01/2019 11:59 am
(@wgdixon)
Posts: 23
Member
 

If I may sort of hijack... if I understand what is happening here I see a scenario in my Mac Pro that may result in (and may have once already) a similar situation. Check me on this.. My Mac Pro rebooted (probably a kernel panic), and when it restarted (which seems to be an issue right now) the two INTERNAL 6GB WD Black drives (which were primary in a 1+0 configuration with two external drives) did not post at all.. never mounted.. so the EXTERNAL drives came up, mounted, and acted pretty much like a good mirrored stripe volume should. So next I powered the Mac Pro down, then back up. The INTERNAL drives and the EXTERNAL drives both mounted as identical striped volumes, missing their mirrors! After reading this I am assuming the EXTERNAL drives failed over and became primary when the INTERNAL drives failed to post, leaving me with two primary drives. Sound like what probably happened? If so I'm pretty concerned because this seems to be an issue with these two internal drives on a restart now will invariably cause me to have a split mirror with no choice but to rebuild when it happens. Any way to force one of these to be secondary or to have a secondary _stay_ secondary on a failover? Seems like a very difficult-to-recover situation if this is the case.

Ok, this worked great, thanks for your help. (Appreciate the tip on pulling the primary.)

What you did was fine, but if you want to move a disk to another computer, move the primary. this is because when you connected the secondary, it failed over. Then SoftRAID cannot know which disk should be the "real" primary.

In your current case, delete the duplicate volume on the ex secondary. then "remove missing secondary disks" and "Add secondary disk" the volume will rebuild.

 
Posted : 08/01/2019 11:51 pm
(@softraid-support)
Posts: 8006
Member Admin
 

You are correct. What you are also encountering is probably a (disk related) TLS issue with the Mac Pro, where the drives may not show up on a restart, but will show up on a cold boot. It happens mostly with larger drives, and I have seen this reported more with WD or the newer Hitachi drives.

Then this behavior is expected.

I cannot think of an easy workaround for this, as I would not want to have the externals be the primary, either.

Normally you could set the Mirror time out to two minutes in case the disks are slow to appear, but in this scenario, the disks will never appear.

 
Posted : 09/01/2019 12:53 am
(@wgdixon)
Posts: 23
Member
 

You are correct. What you are also encountering is probably a (disk related) TLS issue with the Mac Pro, where the drives may not show up on a restart, but will show up on a cold boot. It happens mostly with larger drives, and I have seen this reported more with WD or the newer Hitachi drives.

Then this behavior is expected.

I cannot think of an easy workaround for this, as I would not want to have the externals be the primary, either.

Normally you could set the Mirror time out to two minutes in case the disks are slow to appear, but in this scenario, the disks will never appear.

Thanks. I did some more hunting around in your forums then online about the drives and basically learned the same thing. I did set the timeout to 2 minutes so at least if I'm _there_ if the thing gets knocked over again and wants to reboot, hopefully I can shut it down before things get going. Another thing I found was to set a system parameter that prevents the machine from rebooting on a "freeze".. I don't know if Apple considers a kernel panic to be a "freeze" these days or not..

 
Posted : 09/01/2019 7:58 am
Share:
close
open