Infrastructure
Loadbalancer
For the manage-loadbalancer
play to work, the internal control socket
of the HAProxy service must be set to admin level. As of OSISM 7.0.6 this
is the default. Before this, the parameter haproxy_socket_level_admin
must
be added to the configuration repository and then a reconfigure
(osism apply -a reconfigure loadbalancer
) must be done for the loadbalancer
service.
haproxy_socket_level_admin: "yes"
You can check in the HAProxy configuration whether the control socket is configured correctly.
global
[...]
stats socket /var/lib/kolla/haproxy/haproxy.sock group kolla mode 660 level admin
-
Disable the host
testbed-node-0
in all backends of the servicekeystone
osism apply manage-loadbalancer \
-e manage_loadbalancer_action=disable \
-e manage_loadbalancer_service=keystone \
-e manage_loadbalancer_host=testbed-node-0 -
Enable the host
testbed-node-0
in all backends of the servicekeystone
osism apply manage-loadbalancer \
-e manage_loadbalancer_action=enable \
-e manage_loadbalancer_service=keystone \
-e manage_loadbalancer_host=testbed-node-0 -
Disable the host
testbed-node-0
in all backendsosism apply manage-loadbalancer \
-e manage_loadbalancer_action=disable \
-e manage_loadbalancer_service=all \
-e manage_loadbalancer_host=testbed-node-0 -
Enable the host
testbed-node-0
in all backendsosism apply manage-loadbalancer \
-e manage_loadbalancer_action=enable \
-e manage_loadbalancer_service=all \
-e manage_loadbalancer_host=testbed-node-0
MariaDB
Backup
Mariabackup is used to create backups of MariaDB. For more details about backups, you can use the offical kolla-ansible documentation.
-
Create a full backup
osism apply mariadb_backup
-
Create a incremental backup (supported as of OSISM 7.0.6)
osism apply mariadb_backup -e mariadb_backup_type=incremental
-
Accessing created backups
There is a Docker volume
mariadb_backup
on the 1st control node. The backups are stored in this volume. (see also /var/lib/docker/volumes/mariadb_backup/)$ docker run --rm -v mariadb_backup:/backup -it ubuntu:22.04 bash -c 'ls -la /backup'
total 9728
drwxr-xr-x 2 42434 42434 4096 Jun 3 18:46 .
drwxr-xr-x 1 root root 4096 Jun 3 18:47 ..
-rw-r--r-- 1 42434 42434 4530618 Jun 3 18:46 incremental-18-mysqlbackup-03-06-2024-1717440409.qp.xbc.xbs.gz
-rw-r--r-- 1 42434 42434 11 Jun 3 18:45 last_full_date
-rw-r--r-- 1 42434 42434 5411763 Jun 3 18:45 mysqlbackup-03-06-2024-1717440342.qp.xbc.xbs.gz
Currently there is no offical scheduling and houskeeping (disk space) for mariadb backups. You can create a simple cronjob on the manager or use your enterprise backup software.
cat /etc/cron.d/mariadb_backup <<'EOF'
0 7 * * * dragon osism apply mariadb_backup |logger -t mariadb_backup
EOF
Restore
-
Stop all MariaDb Instances
osism apply -s stop maria
-
Follow the restore procedure described in the kolla-ansible manual
-
Execute the recovery procedure with the node name where you executed the recovery
osism apply mariadb_recovery -e mariadb_recover_inventory_name=THE_NAME_OF_THE_RESTORE_NODE
Recovery
If you stopped your mariadb galera cluster completly, you can use the following procedure to start a recovery.
osism apply mariadb_recovery
Open Search
Get all indices
$ curl https://api-int.testbed.osism.xyz:9200/_cat/indices?v
health status index uuid pri rep docs.count docs.deleted store.size pri.store.size
green open flog-2024.04.17 1rCP3NpUQSS5wmulCn6Y5g 1 1 1657832 0 1gb 654.4mb
green open .opensearch-observability UnS2gFb-QhC8oIefL3C52Q 1 2 0 0 624b 208b
green open .plugins-ml-config hMdzW6ooRMGZ_0OGcdNSgA 1 1 1 0 7.8kb 3.9kb
green open .opendistro-job-scheduler-lock fa_Io8bJQ8qfGII4DypxFg 1 1 1 3 51.1kb 35.1kb
green open .kibana_1 v-aJ6ioSQsOwHQn_NNbeOg 1 1 0 0 416b 208b
Delete an index
$ curl -X DELETE https://api-int.testbed.osism.xyz:9200/flog-2024.04.17
{"acknowledged":true}