[18.11] wal and db LVs are created only on one utility device although multiple are specified
Affects | Status | Importance | Assigned to | Milestone | |
---|---|---|---|---|---|
Ceph OSD Charm |
Triaged
|
Medium
|
Unassigned |
Bug Description
Although the code path in the charm clearly uses find_least_
https:/
and
juju config ceph-osd bluestore-db
/dev/disk/
juju config ceph-osd bluestore-wal
/dev/disk/
I still managed to get into a situation where all wal and db LVs are created on one device only (in this case, an NVMe partition):
tree /dev/disk/by-dname/
/dev/disk/by-dname/
├── md0 -> ../../md0
├── nvme0n1 -> ../../nvme0n1
├── nvme0n1-part1 -> ../../nvme0n1p1
├── nvme0n1-part2 -> ../../nvme0n1p2
├── nvme1n1 -> ../../nvme1n1
├── nvme1n1-part1 -> ../../nvme1n1p1
├── nvme1n1-part2 -> ../../nvme1n1p2
├── osddata0 -> ../../bcache5
├── osddata1 -> ../../bcache3
├── osddata2 -> ../../bcache2
├── osddata3 -> ../../bcache4
├── osddata4 -> ../../bcache0
├── osddata5 -> ../../bcache1
├── sda -> ../../sda
├── sda-part1 -> ../../sda1
├── sdb -> ../../sdb
├── sdc -> ../../sdc
├── sdd -> ../../sdd
├── sde -> ../../sde
├── sdf -> ../../sdf
├── sdf-part1 -> ../../sdf1
├── sdf-part2 -> ../../sdf2
├── sdf-part3 -> ../../sdf3
├── sdg -> ../../sdg
├── sdh -> ../../sdh
└── sdh-part1 -> ../../sdh1
unit log:
2019-03-21 22:53:40 DEBUG secrets-
23e5495d.service → /lib/systemd/
2019-03-21 22:53:40 DEBUG secrets-
2019-03-21 22:53:40 DEBUG secrets-
2019-03-21 22:53:40 DEBUG secrets-
2019-03-21 22:53:40 DEBUG secrets-
2019-03-21 22:53:40 DEBUG secrets-
2019-03-21 22:53:40 DEBUG secrets-
2019-03-21 22:53:40 DEBUG secrets-
2019-03-21 22:53:40 DEBUG secrets-
2019-03-21 22:53:40 DEBUG secrets-
2019-03-21 22:53:40 DEBUG secrets-
2019-03-21 22:53:40 DEBUG secrets-
2019-03-21 22:53:40 DEBUG secrets-
2019-03-21 22:53:40 DEBUG secrets-
2019-03-21 22:53:41 DEBUG secrets-
2019-03-21 22:53:41 DEBUG secrets-
2019-03-21 22:53:41 INFO juju-log secrets-
lsblk
# ...
nvme1n1 259:0 0 2.9T 0 disk
├─nvme1n1p1 259:1 0 676.1G 0 part
│ └─crypt-
│ ├─ceph-
│ ├─ceph-
│ ├─ceph-
│ ├─ceph-
│ ├─ceph-
│ ├─ceph-
│ ├─ceph-
│ ├─ceph-
│ ├─ceph-
│ ├─ceph-
│ ├─ceph-
│ └─ceph-
└─nvme1n1p2 259:2 0 2.1T 0 part
├─bcache0 252:0 0 5.5T 0 disk
│ └─crypt-
│ └─ceph-
├─bcache1 252:128 0 5.5T 0 disk
│ └─crypt-
│ └─ceph-
└─bcache4 252:512 0 5.5T 0 disk
└─crypt-
└
nvme0n1 259:3 0 2.9T 0 disk
├─nvme0n1p1 259:4 0 676.1G 0 part
└─nvme0n1p2 259:5 0 2.3T 0 part
├─bcache2 252:256 0 5.5T 0 disk
│ └─crypt-
│ └─ceph-
├─bcache3 252:384 0 5.5T 0 disk
│ └─crypt-
│ └─ceph-
└─bcache5 252:640 0 5.5T 0 disk
└─crypt-
└
root@block-
LV VG Attr LSize Pool Origin Data% Meta% Move Log Cpy%Sync Convert
osd-block-
osd-block-
osd-block-
osd-block-
osd-block-
osd-block-
osd-db-
osd-db-
osd-db-
osd-db-
osd-db-
osd-db-
osd-wal-
osd-wal-
osd-wal-
osd-wal-
osd-wal-
osd-wal-
summary: |
- [18.11] wal and db LVs are only created on one utility device although + [18.11] wal and db LVs are created only on one utility device although multiple are specified |
TRIAGE:
Interestingly (or not?) the code failed to find the partitions on nvme0 even though they exist; this might be a reason for the failure?