Some disks show up in 'View Disks' when they are actually Multipaths
Dell PowerEdge R715
2 X AMD Opteron 6320
Booting on SD (Dell HW-Mirror)
2 X DNS-1600
48 X 4TB (24 WD + 24 HGST) NearLine SAS-2
Putting on a new shelf/JBOD with 24 new disks and check that the LSI has dualpath to all disks.
FreeNAS is creating 24 new multipath disk - all OK.. But 2 of the 24 disks ( /dev/daxx ) is also showing up in 'View Disks'.
We have seen this before, where the 'solution' was the following:
- sysctl kern.geom.debugflags=0×10
- dd if=/dev/zero of=/dev/daxx bs=512 count=1
- In the GUI choose to 'Wipe' the disk(s) showing both as a single disk AND part of a multipath.
- Reboot FreeNAS
However, this method still gets us the disks in 'View Disks' as if they are singlepath-disks.
Does anyone have a workarround for this or should we just ignore as they seem to be part of a working multipath?
#6 Updated by Josh Paetzel about 5 years ago
- Status changed from Investigation to 15
The GUI is showing the wrong thing because the database is wrong.
The database is wrong because sync_disks is doing the wrong thing somehow. (not taking the multipath members out of the database)
There are a few options here:
1) ignore it, it is just cosmetic and is not affecting anything.
2) Do database surgury to get rid of the cosmetic blemish. I can send you the commands or do a teamviewer with you.
3) Fix the actual bug that is getting this table out of sync in the first place. Once again teamviewer, not recommended for a production system or a system with live data on it for that matter. I'm not perfect and could cause mayhem. Since we don't have anything to repro this and you've run in to it repeatably, there's likely a real bug here. I spent a while introspecting the code and don't see it though.
#7 Updated by Bjarke Emborg Kragelund about 5 years ago
We are in luck as the JBOD and disks are not part of a pool yet. We will isolate the JBOD and disks on a new head. That way we will not be messing with production data.
Can confirm that it very likely is a cosmetic problem. We created a test-pool with all the 24 multipaths - 4 X RaidZ2 vdevs consisting of 6 disks - no glitches. FreeNAS/BSD and ZFS were all happy.
We are two people working on this, and my colleague - for testing purposes - pulled all the disks and inserted them again - the ones from the new JBOD - and all disks now shows up both as Mulitpath and Single. So reproducing the 'error' should be fairly easy.
I will let you know when you can take a look.
As there is some time-difference between us - and there is no data on the system. AND we trust you :-) Maybe a SSH from your IP would suit us all best? Just a thought.
Let me know if that could be solution?