So I went ahead and installed two SSDs in my laptop in order to set up a BtrFS raid. (These SSD’s - samsung 870 - are “blacklisted” so Btrfs shouldn’t try to do anything with them that they can’t handle.) I did the install 3x with similar end results.
I started install… “Expert Partitioner based on current proposal” (which was a single btrfs partition on sda with EFI boot partition and a swapfile). I “edited Btrfs” to “add devices” and added the second SSD (sdb) to the Btrfs with “defaults” for raid. Install went fine (downloading updates during install). But then the system wouldn’t boot (reported no bootable disk). I’ve never seen this behavior before - is there a trick I’m missing here with Btrfs? The setup seems correct with sda having boot/swap partitions and sdb just a Btrfs partition.
In order to boot now, I have to select BIOS esc/F9 (HP ZBook 15 FHD 15.6" Work Station Laptop NoteBook; Core i7-4700MQ; 16GB Ram; Nvidia Quadro K610M) during POST to enter “select boot device” and then select “boot from EFI file” - the BIOS finds the HD and FS and I navigate to the opensuse grub64 EFI file and select that - then the system usually boots correctly. I’m wondering what’s the issue with Btrfs and booting?
The system also periodically slows to a crawl for a few minutes as the btrfs.scrub (etc.) utilities run. I don’t believe Btrfs should slow the system like this! But I guess that’s a separate issue…
Right now my bios is selected to regard “EFI boot + legacy boot” - when I select “EFI only” it won’t boot (but I installed it in “EFI + legacy” mode. I don’t think changing modes is allowed except during initial install.)
Is this an SSD issue? Don’t know. Not enough well-digested information online to diagnose. So I bought two rotating drives and may reinstall to test.
It looks like hybrid SSHDs are super expensive - but I presume these would be faster than just bare rotating drives only. (The things we go through to get a checksumming FS! lol! )
Ideas Appreciated!!
PS: There’s also an issue with Btrfs and “bare metal” SSDs (not sure if it also applies to HDs). If install fails, I had to GParted the drives and create new partition tables (GPT) - that is, take them back to bare metal - b/c YaST install would not deal with the remains of the Btrfs RAID correctly. I got errors associated with missing partitions and unbootability. (Sorry I didn’t document these carefully!)
Thank you very much for the reply! The raid was mostly to avoid dealing with data filling up individual partitions. It’s really nice not to have to move data around - just having one big “partition.” Also makes backing-up very straightforward; I’m “challenged” in this regard, so it’s a big help. I had also thought that the raid would help with the self-error-correction of btrfs but all that is still not clear in my mind - I think the only way that would be possible (with ssd’s) is to have raid1 - a complete mirror ssd.
The booting problem was completely unexpected and not mentioned in any btrfs sites/threads I can find - but I have noticed opensuse slowly getting trickier and trickier to install/update (for instance, lately it has EFI-related issues picking up an existing win7 install and showing win available for boot in GRUB - but that’s a separate laptop and issue…).
All this, of course, is making me wonder about Btrfs…
I didn’t enable legacy boot - “EFI + legacy” mode was the default on this laptop and has been that way since 42.2. (BTW: in the dump below, I had to remove the sandisk to get boot, but I removed it during GRUB, so I guess that’s why it shows up… the system wouldn’t boot from EFI with the sandisk plugged in.)
So…
# parted -l
Model: ATA Samsung SSD 870 (scsi)
Disk /dev/sda: 2000GB
Sector size (logical/physical): 512B/512B
Partition Table: gpt
Disk Flags:
Number Start End Size File system Name Flags
1 1049kB 525MB 524MB fat16 boot, esp
2 525MB 1998GB 1998GB btrfs
3 1998GB 2000GB 2148MB linux-swap(v1) swap
Model: ATA Samsung SSD 870 (scsi)
Disk /dev/sdb: 2000GB
Sector size (logical/physical): 512B/512B
Partition Table: loop
Disk Flags:
Number Start End Size File system Flags
1 0.00B 2000GB 2000GB btrfs
# efibootmgr -v
BootCurrent: 0000
Timeout: 0 seconds
BootOrder: 0000,0001
Boot0000* Notebook Hard Drive BBS(HD,,0x0).......................................................................
Boot0001* USB Hard Drive 1 - SanDisk Ultra Fit BBS(HD,,0x900).......................................................................
Thank you!
I was thinking that I should try reinstalling with “EFI only” in BIOS.
There appear to be issues with SSD’s still reported on the Btrfs homepage(s). It almost reads as if Btrfs development has stalled a bit in that regard.
Rotating drives appears to be the only “guaranteed” way to get self-error-correction with Btrfs (from Btrfs home page).
I did learn about avoiding “secure boot support” (the hard way!). But thanks for mentioning it!
I’ve never seen OpenSuSE self-enable “secure boot support” before! But, oddly, this install with Btrfs raid, it did self-enable it. I had to manually disable it. WEIRD!! (This may be an glitch in the installer code?)
BTW: I’ve been installing/using OS on many machines/servers since OS11 or thereabouts. I’m listed as an “Explorer Penguin” b/c I didn’t update my profile when this forum changed it’s website a year or so ago. (OTOH - I never really became a guru or a wizard or anything like that.)
Thank you very much!! Is it necessary to do a clean reboot then do the efibootmgr thing immediately after?
Hmmm… install went fine - so did the Yast installer get confused? Or did some btrfs.trim/scrub/etc. operation mess something up?
I can’t think of any other possible reasons the efi entry is missing… (I installed OS about 6 days ago.).
It might be related to Yast install since it also has issues with setting up a btrfs raid if the drives have remnants of a prior btrfs raid on them (I wound up doing several repeated installs to get to where I am). I found I had to GParted the drives and create a new GPT on each of them, between each install try, to avoid that.
Double backslashes, huh?
...
-i | --iface name create a netboot entry for the named interface
-l | --loader name (defaults to "\efi\linux\grub.efi")
-L | --label label Boot manager display label (defaults to "Linux")
-m | --mirror-below-4G t|f mirror memory below 4GB
...
Thanks - that’s what I thought, and it explains the double-quotes, I suppose. :shame: (even though the --help didn’t show the escape and it had double-quotes… so confusing!)
And I saw that grubx64.efi was listed as #3 (with that SanDisk as #2). It changed to Notebook Hard Drive on reboot! (but with a couple of failed tries in between)
The failed tries:
tried setting EFI only in BIOS: “No Bootable Image - computer will shut down”
switched back to EFI + legacy
tried booting normally: “No operating system installed…”
tried BIOS F9 boot into “Notebook Hard Drive:” “No operating system installed”
BIOS F9 boot selecting “boot from EFI file” (as I have been doing) (it booted and here I am)
So, I’ll try executing the command again (not sure how to change the boot order with that efibootmgr command) but this time w/o changing bios at all.
… ACTUALLY - now that I think about it - I don’t know if the 870 SSD’s from Samsung are actually black-listed - I think someone told me that the 850’s are. They got that info from a web page that seemed to have some source code listed (or at least the headers). Does anyone know where this information was listed? This could be pretty important b/c if btrfs knows about 850’s but not about 870’s… that could be an issue.
It will be only one in the final output.