Single Host Ceph Server

From HoerupWiki
Revision as of 22:42, 10 November 2020 by Torben (talk | contribs) (Delayed mount)
Jump to: navigation, search

Clean Centos 8

Basic Stuff og cephadm

yum install -y python3 podman chrony lvm2 wget 
wget -O /root/cephadm https://github.com/ceph/ceph/raw/octopus/src/cephadm/cephadm
chmod +x /root/cephadm
mkdir -p /etc/ceph
./cephadm add-repo --release octopus
./cephadm install


Boostrap monitor på egen ip

cephadm bootstrap --mon-ip 192.168.2.206   

Installer ceph

cephadm add-repo --release octopus
cephadm install ceph-common
cephadm install ceph 

Opret OSD'er med alle diske (få lige specifik kommando fra Hoerup)

ceph orch apply osd --all-available-devices
ceph status


Lav ny regel der bruger failure domain på OSD (istedet for 3 hosts)

ceph osd crush rule create-replicated repl1 default osd
ceph osd pool ls
ceph osd pool set device_health_metrics crush_rule repl1


Block Device

EC stuff, her med 4+1

ceph osd pool create rbdmeta replicated repl1
ceph osd erasure-code-profile get default
ceph osd erasure-code-profile set ec41 k=4 m=1 crush-failure-domain=osd
ceph osd pool create rbddata erasure ec41

Hint at denne pool skal bruges til block storage

ceph osd pool application enable rbddata rbd
ceph osd pool application enable rbdmeta rbd

Tillad EC blok overwrites

ceph osd pool set rbddata allow_ec_overwrites true
rbd create --size 40G --data-pool rbddata rbdmeta/ectestimage1
rbd ls rbdmeta

Mapper et rbd image ind som blockdevice

rbd map rbdmeta/ectestimage1

Indskriv i /etc/ceph/rbdmap

rbdmeta/ectestimage1    id=admin,keyring=/etc/ceph/ceph.client.admin.keyring
systemctl enable rbdmap.service


Mount filsystem

mkfs.xfs /dev/rbd0 
mkdir /storage
mount -t xfs /dev/rbd0 /storage/
df -h /storage/

/etc/fstab

/dev/rbd0       /storage/       xfs     defaults,discrad,_netdev        0       0

CephFS - Filesystem

# Filsystemet kalder vi myfs

#setup metadata server
ceph orch apply mds myfs

# opret volume
ceph fs volume create myfs

# metadata OG data pool til rod fs skal være replicated, men vi sætter crushrule for at tillade alle på samme host
ceph osd pool set cephfs.myfs.meta crush_rule repl1
ceph osd pool set cephfs.myfs.data crush_rule repl1

# set intended use på pools
ceph osd pool  application enable  cephfs.myfs.data cephfs
ceph osd pool  application enable  cephfs.myfs.meta cephfs

#brug admin keyring
mount -o name=admin -t ceph 192.168.2.199:/ /mnt/cephfs/

ceph osd pool create cephfs-ec erasure ec41
ceph osd pool  application enable  cephfs-ec cephfs
ceph osd pool set cephfs-ec allow_ec_overwrites true
 
ceph fs add_data_pool myfs cephfs-ec
#create subvolume, utilizing cephfs-ec as backing pool
ceph fs subvolume create myfs subfs --pool_layout cephfs-ec

#mount subvolume, default gets an annoyingly long path
mount -t ceph -o name=admin 192.168.2.199:/volumes/_nogroup/subfs/60918416-6df6-4b4a-a071-ffd527fba26c/ /mnt/cephfs/
#fstab
192.168.2.199:/volumes/_nogroup/subfs/60918416-6df6-4b4a-a071-ffd527fba26c/  /mnt/cephfs ceph  name=admin,_netdev 0 0


Delayed mount

#cat mnt-cephfs.timer
[Unit]
Description=delayed mount of cephfs


[Timer]
OnBootSec=1min
Unit=mnt-cephfs.service



[Install]
WantedBy=timers.target
# cat mnt-cephfs.service
[Unit]
Description=Mount Cephs


[Service]
Type=oneshot
ExecStart=mount /mnt/cephfs
User=root
Group=root

#systemctl enable mnt-cephfs.timer

Hvad mangler vi ?

Clean shutdown / reboot ?

ceph logs ?

Scrubbing ?

Overvågning / prometheus ?

Defekt disk, ny disk.

Rest API

Sources n crap

https://docs.ceph.com/en/latest/cephadm/install/

https://medium.com/@balderscape/setting-up-a-virtual-single-node-ceph-storage-cluster-d86d6a6c658e

https://linoxide.com/linux-how-to/hwto-configure-single-node-ceph-cluster/


Zap disk for re-use

ceph-volume lvm zap /dev/sdX

eller

dd if=/dev/zero of=/dev/vdc bs=1M count=10