Am 21.02.21 um 12:57 schrieb Alexander Wetzel:
While compiling I also switched to
https://github.com/kdave/btrfs-progs.git. Same problem.
I then tracked the error down up to btrfs_uuid_tree_lookup_any():
nr_items is zero after the call
ret = ioctl(fd, BTRFS_IOC_TREE_SEARCH, &search_arg);
(ret is also zero)
So looks like this is a filesystem issue?
I've now "reused" the ret < 0 output in btrfs_uuid_tree_lookup_any() and
added nt_items to it, too.
Then I get:
# /home/alex/src/b2/btrfs-progs/btrfs receive -f test2 .
At snapshot 2021-02-20-TEMP
ioctl(BTRFS_IOC_TREE_SEARCH, uuid, key 483c17093f551dd3, UUID_KEY,
896d0cfd51ec5cb6, nr_items=0) ret=0, error: No such file or directory
ioctl(BTRFS_IOC_TREE_SEARCH, uuid, key 483c17093f551dd3, UUID_KEY,
896d0cfd51ec5cb6, nr_items=0) ret=0, error: No such file or directory
ERROR: clone: did not find source subvol si=0,
uuid=d31d553f-0917-3c48-b65c-ec51fd0c6d89
In common/send-stream.c read_and_process_cmd() the very first time we
try to use BTRFS_SEND_C_CLONE it triggers the problem on my setup.
Dumping the variables of the problematic call here I get:
path=var/lib/mysql/nextcloud/oc_activity.ibd
offset=745472
len=303104
uuid=d31d553f-0917-3c48-b65c-ec51fd0c6d89
clone_ctransid=195056
clone_path=var/lib/mysql/nextcloud/oc_activity.ibd
clone_offset=745472
Guessing around I assume this is an instruction to clone 303104 Bytes at
offset 745472 for var/lib/mysql/nextcloud/oc_activity.ibd from the
parent subvol uuid d31d553f-0917-3c48-b65c-ec51fd0c6d89 at generation
195056.
So I suspect now my problem is, that my RO snapshot hs somehow
Generation != (Gen at creation).
A RO snapshot should not be able to get to a newer generation, correct?
Or maybe that btrfs send is using "Gen at creation" instead of
"Generation"... but don't think so.
# btrfs sub show 2021-02-14/
2021-02-14
Name: 2021-02-14
UUID: d31d553f-0917-3c48-b65c-ec51fd0c6d89
Parent UUID: a2ff485b-9f8f-2543-ad03-dc7c917e143b
Received UUID: -
Creation time: 2021-02-14 21:46:26 +0100
Subvolume ID: 358
Generation: 208977 <---
Gen at creation: 195056
Parent ID: 5
Top level ID: 5
Flags: readonly
Snapshot(s):
2021-02-20-TEMP
Now I somehow got another test ro snapshot to a new generation:
# btrfs sub snap -r active/ t1
Create a readonly snapshot of 'active/' in './t1'
# btrfs sub show t1
t1
Name: t1
UUID: 70e6a915-e03a-fb48-869e-98ab8bad2ad4
Parent UUID: a2ff485b-9f8f-2543-ad03-dc7c917e143b
Received UUID: -
Creation time: 2021-02-21 18:02:32 +0100
Subvolume ID: 411
Generation: 209009
Gen at creation: 209009
Parent ID: 5
Top level ID: 5
Flags: readonly
Snapshot(s):
Playing around a bit I found out that renaming a subvol can - but not
always - advances the generation:
# btrfs sub snap -r active/ t1
Create a readonly snapshot of 'active/' in './t1'
# btrfs sub show t1
t1
Name: t1
UUID: 83d278b0-1782-3a4a-8293-c6df2e67acd4
Parent UUID: a2ff485b-9f8f-2543-ad03-dc7c917e143b
Received UUID: -
Creation time: 2021-02-21 18:34:40 +0100
Subvolume ID: 414
Generation: 209069
Gen at creation: 209069
Parent ID: 5
Top level ID: 5
Flags: readonly
Snapshot(s):
# mv t1 x1
# btrfs sub show x1
x1
Name: x1
UUID: 83d278b0-1782-3a4a-8293-c6df2e67acd4
Parent UUID: a2ff485b-9f8f-2543-ad03-dc7c917e143b
Received UUID: -
Creation time: 2021-02-21 18:34:40 +0100
Subvolume ID: 414
Generation: 209070
Gen at creation: 209069
Parent ID: 5
Top level ID: 5
Flags: readonly
Snapshot(s):
But that is not always working. Creating/renaming/deleting snapshots I
can trigger it from time to time but so far I don't see a pattern.
But then I just got that:
# btrfs sub snap -r active/ q1
Create a readonly snapshot of 'active/' in './q1'
# btrfs sub show q1
q1
Name: q1
UUID: 53a3833c-9caa-b944-80d1-8d592c68ff74
Parent UUID: a2ff485b-9f8f-2543-ad03-dc7c917e143b
Received UUID: -
Creation time: 2021-02-21 18:40:20 +0100
Subvolume ID: 416
Generation: 209084
Gen at creation: 209083
Parent ID: 5
Top level ID: 5
Flags: readonly
Snapshot(s):
root@mail:/mnt/backup/backupdev#
No rename necessary and the generations also differ...
But I guess I'm at then end of what I can find out without a much better
understanding of btrfs internals...
I hope someone here can shed more light on that,
Alexander
