Actions
Bug #6966
closedceph-disk: prepare --dmcrypt failing
% Done:
0%
Source:
Support
Tags:
Backport:
dumpling
Regression:
Severity:
3 - minor
Reviewed:
Affected Versions:
ceph-qa-suite:
Pull request ID:
Crash signature (v1):
Crash signature (v2):
Description
# ceph-disk prepare --dmcrypt /dev/sdh INFO:ceph-disk:Will colocate journal with data on /dev/sdh Information: Moved requested sector from 34 to 2048 in order to align on 2048-sector boundaries. The operation has completed successfully. Information: Moved requested sector from 10485761 to 10487808 in order to align on 2048-sector boundaries. Warning: The kernel is still using the old partition table. The new table will be used at the next reboot. The operation has completed successfully. ceph-disk: Error: partition 1 for /dev/sdh does not appear to exist
Looks like this is due to partprobe not being ran after creating the data partition.
This patch seems to do the trick.
diff --git a/src/ceph-disk b/src/ceph-disk
index 0691d25..df615dc 100755
--- a/src/ceph-disk
+++ b/src/ceph-disk
@@ -1044,6 +1044,12 @@ def prepare_dev(
)
subprocess.call(
args=[
+ 'partprobe',
+ data,
+ ],
+ )
+ subprocess.call(
+ args=[
# wait for udev event queue to clear
'udevadm',
'settle',
Updated by Eric Eastman almost 10 years ago
This is still failing in Firefly 0.80-rc1
# ceph -v ceph version 0.80-rc1-33-g0f3235d (0f3235d46c8fd6c537bd4aa8a3faec6c00f311a8) # ceph-disk prepare --dmcrypt /dev/sdc INFO:ceph-disk:Will colocate journal with data on /dev/sdc The operation has completed successfully. Warning: The kernel is still using the old partition table. The new table will be used at the next reboot. The operation has completed successfully. ceph-disk: Error: partition 1 for /dev/sdc does not appear to exist
I applied the above patch, with the following modifications for Firefly:
diff -c5 /usr/sbin/ceph-disk.org /usr/sbin/ceph-disk
*** /usr/sbin/ceph-disk.org 2014-05-01 10:39:29.782630806 -0600
--- /usr/sbin/ceph-disk 2014-05-01 10:40:40.944066600 -0600
***************
*** 1197,1206 ****
--- 1197,1212 ----
data,
],
)
command(
[
+ 'partprobe',
+ data,
+ ],
+ )
+ command(
+ [
# wait for udev event queue to clear
'udevadm',
'settle',
],
)
Using the patched version, it now works:
# ceph-disk prepare --dmcrypt /dev/sdc
INFO:ceph-disk:Will colocate journal with data on /dev/sdc
The operation has completed successfully.
Warning: The kernel is still using the old partition table.
The new table will be used at the next reboot.
The operation has completed successfully.
meta-data=/dev/mapper/c4c38377-e2df-46f4-98ca-928115adf650 isize=2048 agcount=4, agsize=2293695 blks
= sectsz=512 attr=2, projid32bit=0
data = bsize=4096 blocks=9174779, imaxpct=25
= sunit=0 swidth=0 blks
naming =version 2 bsize=4096 ascii-ci=0
log =internal log bsize=4096 blocks=4479, version=2
= sectsz=512 sunit=0 blks, lazy-count=1
realtime =none extsz=4096 blocks=0, rtextents=0
Warning: The kernel is still using the old partition table.
The new table will be used at the next reboot.
The operation has completed successfully.
Updated by Sage Weil almost 10 years ago
- Status changed from New to Fix Under Review
Updated by Eric Eastman almost 10 years ago
I installed wip-6966 on a test cluster and was able to build and use dmcrypted OSDs.
Thanks!
Updated by Alfredo Deza almost 10 years ago
- Status changed from Fix Under Review to Resolved
Merged to master with hash 0f196265f049d432e399197a3af3f90d2e916275
Updated by Sage Weil almost 10 years ago
- Status changed from Resolved to Pending Backport
Updated by Sage Weil almost 10 years ago
- Status changed from Pending Backport to Resolved
- Backport set to dumpling
Actions