answered
2016-03-03 11:14:51 +0200
For reference, it looks like this is what they did on the tablet. From the screenshots I've seen of the Intex phone, it also has two separate partitions which hints that it may be the same. Btrfs was a bold experiment, but this is much better. Now if there were only some way that we could create a similar image for the Jolla phone...
lvm> lvdisplay
--- Logical volume ---
LV Path /dev/sailfish/root
LV Name root
VG Name sailfish
LV UUID tYDgK9-HJM1-4c3Q-4Mu5-j9QK-EJz6-crACDt
LV Write Access read/write
LV Creation host, time (none), 2015-09-01 17:25:44 +0200
LV Status available
# open 1
LV Size 3.91 GiB
Current LE 1000
Segments 1
Allocation inherit
Read ahead sectors auto
- currently set to 256
Block device 253:0
--- Logical volume ---
LV Path /dev/sailfish/home
LV Name home
VG Name sailfish
LV UUID LTQDU5-Zny0-Xg7p-svS9-GPV6-P4jD-VciSLr
LV Write Access read/write
LV Creation host, time (none), 2015-09-01 17:25:44 +0200
LV Status available
# open 1
LV Size 49.98 GiB
Current LE 12795
Segments 1
Allocation inherit
Read ahead sectors auto
- currently set to 256
Block device 253:1
btrfs is the future, and it is only going to get better, what you propose is a bit ludicrous!
dsilveira ( 2014-11-24 21:30:52 +0200 )editWell, if I understand the discussion correctly, btrfs might have some inherent design flaws which can not be corrected so easily. Those leads to really bad disk utilization so that e.g. a disk filled half with data is to be considered full. The filesystem b-tree is then extremely unbalanced. I don't know whether this can sanely be changed without dirty hacks or changing the filesystem specification, which would mean introducing a btrfs2. That probably others have to investigate.
Balancing seems to be a method to get around this, but balancing the btrfs filesystem as far as I understand involves a lot of writes, which is not optimal for the flash memory. Even without rebalancing, according to some presentations I saw btrfs does already more write operations than ext4 with journaling disabled. So the question would be, is there a solution for the problems with btrfs in the very near future? If not, I think switching to ext4 could be an option. It makes no sense to stay now with a filesystem that has some serious issues and annoying problems just because we assume that some day in the future it will work. Then lets switch to ext4 and consider btrfs again when it is stable.
dac ( 2014-11-25 13:18:10 +0200 )editI have no idea what you're talking about, I've never heard any of that. As for stability, I use it everyday on my phone, laptop, and server, no issue, no instability! Distros such as fedora and others are going to make it default for the next release, so what have you been smoking lately?
dsilveira ( 2014-11-25 21:10:15 +0200 )editApart from the personal attack, to which part of my posting to you refer with your statements?
Which part of what I wrote you have no idea about?
B-Trees and their general characteristics
The claimed "design flaws", about which I am not sure I can make a definitive assessment myself
The fact that there are people using Jolla who have this kind of errors, as described in the link provided in the question?
The fact that when you do a balance on btrfs it means writing a significant amount of data?
The idea that if there is some serious enough issue with some component you are using in a system in a specific application, and if there is no short term fix, you might think about alternatives bridging this time until a fix is available?
About balancing the trees, yes there is the idea that the tree in btrfs is self balancing, and actually a good implementation would be just that. I just remember the analysis from Shishkin where he showed that in btrfs it is not. I am not sure whether that has changed... But be it so.
Balance operation according to kernel.org means rewriting all data and metadata again on the disk, which is essentially even worse for a flash drive than a simple "rebalancing" of the b-trees used.
I do not know whether the criticism of Shishkin has been answered or falsified by someone, at least I don't remember seeing that. I do not know whether the situation that he described has been altered with a fix. I only saw that there are people having this issue in their Jolla, so obviously something could need a fix. And I saw there and elsewhere that repairing the situation is not really straightforward. I remember for desktops someone hinted that one should add a usb drive to the btrfs volume so that enough space is available for balance. This is obviously difficult using Jolla, though one could use the microsd for it also.
dac ( 2014-11-25 23:37:39 +0200 )editAbout the "the others do it" argument:
First, "doing what others do just because others do" is not always a good idea just without any additional sane reasoning
Second, this reasoning is the mirror of what I read in the beginning of the year relating to Fedora, when people asked to finally make btrfs default for Fedora because Sailfish OS and others are using it already. Circular reasoning...
Third, Fedora is discussing this already quite a while (it is planned to switch since 2011), and there are people involved who seem to think that btrfs is not ready yet for this. And it seems that it might not happen in the next release yet, but for Fedora 23 or later, which should come end of next year. And I am not sure if that discussion is over already. And it was specifically mentioned that although f.ex. OpenSuse uses it as default already, one of the reasons seem to be that "Suse benefits from having a narrow "supported" criteria, like only use it with lots of space....", see here. This "lots of space" does obviously not mean 16GB.
And anecdotal experience like yours is nice, but it does not help those who encounter this issue. I could also claim I cross streets every day on red traffic lights and have no issues with that. It would not mean anything about generalization or about having that as a sane default. Not that I do....
What I wanted to express in my comment was that IF there seems to be stability issues with btrfs on Sailfish OS which lead to serious problems, which seems plausible according to the question linked, one possibility COULD be switching to another FS that does have better characteristics in the specific application. I think that is fair enough, and I am looking forward to reasoning and facts falsifying my statements and opinion, but please try to do it using your brain and not with personal attacks. After all this is a platform for discussion.
dac ( 2014-11-25 23:40:38 +0200 )edit