[tlug] linuxcon2k15 thoughts & topics, 1

[Christian Horn <chorn@example.com>]

Hoi,

random ideas and thoughts, in the hope that others on
the list share their thinking regarding the topics.

# dif/dix
I got aware of the topic again and was looking up the state.
Overall issue to be solved: imagine you run an application 
on linux, in a virtual guest, and write data onto a file-
system on SAN.
At some point the application complains about corrupted data.
Which layer is it?  How could one better pin this down?
dif/dix provides checksums over data written, the application
creates them, they traverse the layers and in the end the 
storage verifies.
State: Red Hat Enterprise Linux (RHEL) includes the code, but
the feature has techpreview state, so not meant for real pro-
duction data, so not fully supported.
Oracle supports it already.
Interesting feature.  So far, XFS is on filesystem layer only
offering metadata checksums.
Further details: https://access.redhat.com/solutions/41548

# btrfs
Red Hat does not support it.  We have coders working on it,
put many efforts, but consider it not production ready. 
The code is shipped as techpreview, bugreports are helpful.
Suse and oracle confirmed they support it (with restrictions 
to certain configs/raidlevels).
At the kernel symposium discussion I asked if there would be
a good way to break out of the loop of not many users/the code
this not beeing widely tested/horror stories on dataloss con-
tinuing.  According to the experts (Chris Mason and others) 
the situation here is slowly improving, code stabilizing.
The features of btrfs can partly be done in other layers,
i.e. lvm snapshots.

Enough for today, should become 3-5 of such mails.
All thoughts welcome.
Chris