Well I focused on btrfs because you mentioned it as one of the options…
Moreover, I checked here: btrfs.wiki.kernel.org/index.php … un_2012.29 and kernel 3.18 is already very fine for it (which devices cannot be updated?).
In case you change your mind, this is a step by step to convert ext4 images to btrfs on R-Pi, usable also on other platforms: raspberrypi.org/forums/view … 9&t=104108
That said, let’s try SquasFS.
Still, may I very warmly ask to add a fixed delay (in the rc shutdown scripts) between final unmounting of the filesystems and actual power off? SD cards may need some seconds, the ones that do wear leveling by themselves, and they do not report any “ready” or “completed” status. If you power down quickly… problems may arise.
I think 5-10 seconds are enough (and should not bring any additional issue).
I think you have to edit the rc0.d (runlevel 0, shutdown) init files in /etc with a “sleep” command somewhere. unix.stackexchange.com/questions … ies-in-etc
Absolutely agree…
But, we have the squashfs images written on a ext4 partition (with no journaling), do you think that this would not be sufficient to avoid corruption?
I would say lets try it. Even though changes to the file system would require re-flashing the SD cards it some thing most people can live with. If Volumio 2 is out in the open there will always be people who would’ve rather seen an different file-system, but then they can try and see if there is one that works better than the original delivered version. After that and some hard numbers the change should be possible via a new image
Yes, let’s try and see. I may try myself to convert to btrfs and provide a guide later on.
Only, why disabling journaling on ext4? Volumio is not a write-intensive server, journaling would not reduce performances but only give additional integrity checks.
I think I read on stackoverflow the journaling writes a lot of data and therefore is not good for sd-cards, because they may get destortyed. But I have no evidence
raspberrypi.stackexchange.com/qu … ng-enabled
Right, I see. It implicitly confirms that wear leveling is not performed by all cards (otherwise the sector with the journal would not wear out before the rest of the card).
I always saw that to prevent SD Card wear journaling should be disabled… But I admit I also have no evidence to support that.
For the image, a bit of patience more… All the pieces are coming together…
Yes, I’ll do… But really that unfortunately gets a low priority on my scale… I’m so eager to see Volumio2 running that I’m putting every moment I have into coding…