#999 Issue closed: since SLES12-SP2 btrfs quota via "snapper setup-quota" is needed¶
Labels: enhancement, fixed / solved / done
jsmeix opened issue at 2016-09-16 08:14:¶
After
https://github.com/rear/rear/issues/944
and
https://github.com/rear/rear/issues/963
which was in the end a false-positive
I finally found (by luck) a real change from SLES12-SP1 to SP2
that actually requires an appropriate enhancement
in Relax-and-Recover:
Since SLES12-SP2 btrfs quota via "snapper setup-quota" is used
which means "snapper setup-quota" must be called
during "rear recover" if that was used in the original system.
I.e. during "rear mkbackup" I need to autodetect whether or not
btrfs quota via "snapper setup-quota" is used and if yes store that
info somehow in disklayout.conf (I assume filesystem quota info
belongs to disklayout.conf) so that "rear recover" could call
"snapper setup-quota" when needed.
Regarding how btrfs quota is used for snapper see
http://snapper.io/2016/05/18/space-aware-cleanup.html
For the reasoning behind why btrfs quota is used for snapper see
https://lists.opensuse.org/opensuse-factory/2016-09/msg00074.html
https://lists.opensuse.org/opensuse-factory/2016-09/msg00121.html
https://lists.opensuse.org/opensuse-factory/2016-09/msg00124.html
For details probably best is to read the whole (very interesting)
"btrfs quotas enabled by default" mail thread in Sep 2016
on the opensuse-factory@opensuse.org mailing list.
jsmeix commented at 2016-09-16 08:52:¶
This issue is no bug in Relax-and-Recover because
a)
missing support for newest stuff in newest Linux distribution versions
is never ever a bug (but it is an enhancement), and
b)
current "rear recover" (without support for "snapper setup-quota")
works sufficiently well.
After current "rear recover" the admin must manually re-setup
btrfs quota for snapper in the recovered system according to
what is described under "steps for manually setup" in
http://snapper.io/2016/05/18/space-aware-cleanup.html
How I did manual btrfs quota re-setup for snapper
in a recovered SLES12-SP2 system
(I run "yast users" only as an example to get some snapshots):
# btrfs qgroup show -p / ERROR: can't perform the search - No such file or directory ERROR: can't list qgroups: No such file or directory # snapper setup-quota Quota failure (qgroup already set). # snapper get-config | grep QGROUP QGROUP | 1/0 # snapper set-config QGROUP= # snapper get-config | grep QGROUP QGROUP | # snapper setup-quota # btrfs qgroup show -p / qgroupid rfer excl parent -------- ---- ---- ------ 0/5 16.00KiB 16.00KiB --- ... 0/278 16.00KiB 16.00KiB --- 0/279 3.15GiB 3.15GiB --- 1/0 0.00B 0.00B --- # snapper list Type | # | Pre # | Date | User | Cleanup | Description | Userdata -------+---+-------+-------------------------------+------+---------+-----------------------+--------- single | 0 | | | root | | current | single | 1 | | Fri 16 Sep 2016 10:27:55 CEST | root | | first root filesystem | # yast users # snapper list Type | # | Pre # | Date | User | Cleanup | Description | Userdata -------+---+-------+-------------------------------+------+---------+-----------------------+--------- single | 0 | | | root | | current | single | 1 | | Fri 16 Sep 2016 10:27:55 CEST | root | | first root filesystem | pre | 2 | | Fri 16 Sep 2016 10:33:04 CEST | root | number | yast users | post | 3 | 2 | Fri 16 Sep 2016 10:33:10 CEST | root | number | | # btrfs subvolume list / | grep snapshots ID 278 gen 51 top level 257 path @/.snapshots ID 279 gen 49 top level 278 path @/.snapshots/1/snapshot ID 284 gen 48 top level 278 path @/.snapshots/2/snapshot ID 285 gen 49 top level 278 path @/.snapshots/3/snapshot # btrfs qgroup show -p / qgroupid rfer excl parent -------- ---- ---- ------ 0/5 16.00KiB 16.00KiB --- ... 0/278 16.00KiB 16.00KiB --- 0/279 3.15GiB 16.00KiB --- 0/284 3.15GiB 3.16MiB 1/0 0/285 3.15GiB 16.00KiB 1/0 1/0 0.00B 3.14MiB --- # btrfs quota rescan / quota rescan started # btrfs quota rescan -s / no rescan operation in progress # btrfs qgroup show -p / qgroupid rfer excl parent -------- ---- ---- ------ 0/5 16.00KiB 16.00KiB --- ... 0/278 16.00KiB 16.00KiB --- 0/279 3.15GiB 16.00KiB --- 0/284 3.15GiB 3.16MiB 1/0 0/285 3.15GiB 16.00KiB 1/0 1/0 3.15GiB 3.17MiB ---
The last "btrfs qgroup show -p /" here matches that output
in
http://snapper.io/2016/05/18/space-aware-cleanup.html
so that my manual btrfs quota setup for snapper looks o.k.
jsmeix commented at 2016-09-16 10:25:¶
I tried to automate it with
(long line shown wrapped here):
POST_RECOVERY_SCRIPT=( '/usr/bin/snapper --no-dbus -r /mnt/local set-config QGROUP= && echo snapper set empty QGROUP succeeded || echo snapper set empty QGROUP failed' '/usr/bin/snapper --no-dbus -r /mnt/local setup-quota && echo snapper setup-quota succeeded || echo snapper setup-quota failed' )
in /etc/rear/local.conf but that does not work because
usr/share/rear/wrapup/default/50_post_recovery_script.sh
runs
eval "${POST_RECOVERY_SCRIPT[@]}"
instead of
for command in "${POST_RECOVERY_SCRIPT[@]}" ; do
eval "$command"
done
but after I also changed 50_post_recovery_script.sh this way
I have automated btrfs quota re-setup for snapper
during "rear recover" and in the recovered system I have now
(I run "yast users" only as an example to get some snapshots):
# btrfs qgroup show -p / qgroupid rfer excl parent -------- ---- ---- ------ 0/5 16.00KiB 16.00KiB --- ... 0/278 16.00KiB 16.00KiB --- 0/279 3.15GiB 3.15GiB --- 1/0 0.00B 0.00B --- 255/260 24.00KiB 24.00KiB --- # yast users # snapper list Type | # | Pre # | Date | User | Cleanup | Description | Userdata -------+---+-------+-------------------------------+------+---------+-----------------------+--------- single | 0 | | | root | | current | single | 1 | | Fri 16 Sep 2016 11:49:56 CEST | root | | first root filesystem | pre | 2 | | Fri 16 Sep 2016 12:12:40 CEST | root | number | yast users | post | 3 | 2 | Fri 16 Sep 2016 12:12:47 CEST | root | number | | # btrfs subvolume list / | grep snapshots ID 278 gen 45 top level 257 path @/.snapshots ID 279 gen 45 top level 278 path @/.snapshots/1/snapshot ID 284 gen 44 top level 278 path @/.snapshots/2/snapshot ID 285 gen 45 top level 278 path @/.snapshots/3/snapshot # btrfs qgroup show -p / qgroupid rfer excl parent -------- ---- ---- ------ 0/5 16.00KiB 16.00KiB --- ... 0/278 16.00KiB 16.00KiB --- 0/279 3.15GiB 16.00KiB --- 0/284 3.15GiB 3.16MiB 1/0 0/285 3.15GiB 16.00KiB 1/0 1/0 0.00B 3.14MiB --- 255/260 24.00KiB 24.00KiB ---
All what is missing is "btrfs quota rescan /"
to get for the "1/0" btrfs qgroup the actually
right amount of "rfer" (referenced size).
I did that manually:
# btrfs quota rescan / quota rescan started # btrfs qgroup show -p / | grep '1/0' ... 1/0 3.15GiB 3.45MiB --- ...
jsmeix commented at 2016-09-16 12:50:¶
It seems "btrfs quota rescan /" does not work reliably
(or does not show reliably results):
I did one more recovery and now in the recovered system
I did several times "btrfs quota rescan /" without an effect.
After each "btrfs quota rescan /" I waited some time
but then I got several times
# btrfs qgroup show -p / ... 1/0 0.00B 0.00B --- ...
that changed after some time into
# btrfs qgroup show -p / ... 1/0 0.00B 3.17MiB --- ...
until finally it became
# btrfs quota rescan / quota rescan started # btrfs qgroup show -p / ... 1/0 3.15GiB 3.50MiB --- ...
Of course I had also tried "btrfs quota rescan -w /"
and "btrfs quota rescan -s /" but I never got a respose
that a rescan operation would be currently still running
(i.e. for me it seems any rescan operation was basically
immediately finished).
jsmeix commented at 2016-09-16 13:37:¶
To get during "rear recover" automated btrfs quota
re-setup for snapper if that is used in the original system
the following works for me in /etc/rear/local.conf
(a single long line shown wrapped here):
POST_RECOVERY_SCRIPT=( '/usr/bin/snapper --no-dbus -r /mnt/local get-config | grep -q "^QGROUP.*1/0" && ( echo snapper setup-quota && /usr/bin/snapper --no-dbus -r /mnt/local set-config QGROUP= && /usr/bin/snapper --no-dbus -r /mnt/local setup-quota || echo snapper setup-quota failed ) || echo snapper setup-quota not used' )
without any additional adaptions in
usr/share/rear/wrapup/default/50_post_recovery_script.sh
(and without "btrfs quota rescan" support).
jsmeix commented at 2016-09-20 11:41:¶
I asked our (i.e. SUSE's) snapper expert and according to him
running "snapper setup-quota" alone is sufficient to setup btrfs quota
for snapper (i.e. no additional "btrfs quota rescan /" needed).
Therefore the POST_RECOVERY_SCRIPT in the current
SLE12-SP2-btrfs-example.conf is sufficient to fix the issue.
I tested it on SLES12-SP2-RC2 with its default btrfs structure
and for me it "just works".
[Export of Github issue for rear/rear.]