1. Installation 2. Devstack Plugin 3. Concepts and definitions 4. API registration 5. API routes 6. Backup metadata structure 7. Freezer Client document structure 8. Jobs 9. Actions 10. Sessions
# git clone https://git.openstack.org/openstack/freezer-api.git # cd freezer-api && sudo python setup.py install
# pip install -r requirements.txt
# sudo cp etc/freezer-api.conf /etc/freezer-api.conf # sudo vi /etc/freezer-api.conf
The currently supported db is Elasticsearch. In case you are using a dedicated instance of the server, you'll need to start it. Depending on the OS flavor it might be a:
# service elasticsearch start
or, on systemd:
# systemctl start elasticsearch
Elasticsearch needs to know what type of data each document's field contains. This information is contained in the "mapping", or schema definition. Elasticsearch will use dynamic mapping to try to guess the field type from the basic datatypes available in JSON, but some field's properties have to be explicitly declared to tune the indexing engine. To do that, use the freezer-db-init command:
# freezer-db-init [db-host]
The url of the db-host is optional and can be automatically guessed from /etc/freezer-api.conf
To get information about optional additional parameters:
freezer-db-init -h
# freezer-api
# uwsgi --http :9090 --need-app --master --module freezer_api.cmd.api:application # uwsgi --https :9090,foobar.crt,foobar.key --need-app --master --module freezer_api.cmd.api:application
To configure the Freezer API with DevStack, you will need to enable the freezer-api plugin by adding one line to the [[local|localrc]] section of your local.conf file:
enable_plugin freezer-api <GITURL> [GITREF]
where:
<GITURL> is the URL of a freezer-api repository [GITREF] is an optional git ref (branch/ref/tag). The default is master.
For example:
enable_plugin freezer-api https://git.openstack.org/openstack/freezer-api.git master
The plugin makes use of apache2 by default. To use the uwsgi server set the following environment variable:
export FREEZER_API_SERVER_TYPE=uwsgi
The default port is 9090. To configure the api to listen on a different port set the variable FREEZER_API_PORT. For example to make use of port 19090 use:
export FREEZER_API_PORT=19090
For more information, see: http://docs.openstack.org/developer/devstack/plugins.html
hostname is _probably_ going to be the host fqdn.
backup_id defined as "container_hostname_backupname_timestamp_level" uniquely identifies a backup
backup_set defined as "container_hostname_backupname" identifies a group of related backups which share the same container,hostname and backupname
# keystone user-create --name freezer --pass FREEZER_PWD
# keystone user-role-add --user freezer --tenant service --role admin
# keystone service-create --name freezer --type backup \
--description "Freezer Backup Service"
# keystone endpoint-create \
--service-id $(keystone service-list | awk '/ backup / {print $2}') \
--publicurl http://freezer_api_publicurl:port \
--internalurl http://freezer_api_internalurl:port \
--adminurl http://freezer_api_adminurl:port \
--region regionOne
GET / List API version GET /v1 JSON Home document, see http://tools.ietf.org/html/draft-nottingham-json-home-03
GET /v1/backups(?limit,marker) Lists backups
POST /v1/backups Creates backup entry
GET /v1/backups/{backup_id} Get backup details
UPDATE /v1/backups/{backup_id} Updates the specified backup
DELETE /v1/backups/{backup_id} Deletes the specified backup
GET /v1/clients(?limit,offset) Lists registered clients
POST /v1/clients Creates client entry
GET /v1/clients/{freezerc_id} Get client details
UPDATE /v1/clients/{freezerc_id} Updates the specified client information
DELETE /v1/clients/{freezerc_id} Deletes the specified client information
GET /v1/jobs(?limit,offset) Lists registered jobs
POST /v1/jobs Creates job entry
GET /v1/jobs/{jobs_id} Get job details
POST /v1/jobs/{jobs_id} creates or replaces a job entry using the specified job_id
DELETE /v1/jobs/{jobs_id} Deletes the specified job information
PATCH /v1/jobs/{jobs_id} Updates part of the document
GET /v1/actions(?limit,offset) Lists registered action
POST /v1/actions Creates action entry
GET /v1/actions/{actions_id} Get action details
POST /v1/actions/{actions_id} creates or replaces a action entry using the specified action_id
DELETE /v1/actions/{actions_id} Deletes the specified action information
PATCH /v1/actions/{actions_id} Updates part of the action document
GET /v1/sessions(?limit,offset) Lists registered session
POST /v1/sessions Creates session entry
GET /v1/sessions/{sessions_id} Get session details
POST /v1/sessions/{sessions_id} creates or replaces a session entry using the specified session_id
DELETE /v1/sessions/{sessions_id} Deletes the specified session information
PATCH /v1/sessions/{sessions_id} Updates part of the session document
POST /v1/sessions/{sessions_id}/action requests actions (e.g. start/end) upon a specific session
PUT /v1/sessions/{sessions_id}/jobs/{job_id} adds the job to the session
DELETE /v1/sessions/{sessions_id}/jobs/{job_id} adds the job to the session
NOTE: sizes are in MB
backup_metadata:=
{
"container": string,
"host_name": string, # fqdn, client has to provide consistent information here !
"backup_name": string,
"timestamp": int,
"level": int,
"max_level": int,
"mode" : string, (fs mongo mysql)
"fs_real_path": string,
"vol_snap_path": string,
"total_broken_links" : int,
"total_fs_files" : int,
"total_directories" : int,
"backup_size_uncompressed" : int,
"backup_size_compressed" : int,
"compression_alg": string, (gzip bzip xz)
"encrypted": bool,
"client_os": string
"broken_links" : [string, string, string],
"excluded_files" : [string, string, string]
"cli": string, equivalent cli used when executing the backup ?
"version": string
}
The api wraps backup_metadata dictionary with some additional information. It stores and returns the information provided in this form:
{
"backup_id": string # container_hostname_backupname_timestamp_level
"user_id": string, # owner of the backup metadata (OS X-User-Id, keystone provided)
"user_name": string # owner of the backup metadata (OS X-User-Name, keystone provided)
"backup_metadata": { #--- actual backup_metadata provided
"container": string,
"host_name": string,
"backup_name": string,
"timestamp": int,
...
}
}
Identifies a freezer client for the purpose of sending action
client_info document contains information relevant for client identification:
client_info:=
{
"client_id": string actually a concatenation "tenant-id_hostname"
"hostname": string
"description": string
"uuid":
}
client_type document embeds the client_info and adds user_id:
client_type :=
{
"client" : client_info document,
"user_id": string, # owner of the information (OS X-User-Id, keystone provided, added by api)
}
A job describes a single action to be executed by a freezer client, for example a backup, or a restore. It contains the necessary information as if they were provided on the command line.
A job is stored in the api together with some metadata information such as: job_id, user_id, client_id, status, scheduling information etc
Scheduling information enables future/recurrent execution of jobs
+---------------------+
| Job |
+---------------------+ job_actions +--------------+
| +---------------->| job_action |
| +job_id | 0..* +--------------+ freezer_action
| +client_id | | +mandatory |-------------+
| +user_id | | +retries | | +----------------+
| +description | job_schedule +--------------+ +->| freezer_action |
| +---------------+ +----------------+
| | | +-------------------+
+---------------------+ +-->| job schedule dict |
+-------------------+
job document structure:
"job": {
"job_action": { parameters for freezer to execute a specific action }
"job_schedule": { scheduling information }
"job_id": string
"client_id": string
"user_id": string
"description": string
}
"job_actions":
[
{
"freezer_action" :
{
"action" : string
"mode" : string
"src_file" : string
"backup_name" : string
"container" : string
...
},
"mandatory": False,
"max_retries": 3,
"max_retry_interval": 60
},
{
"freezer_action" :
{
...
},
"mandatory": False,
"max_retries": 3,
"max_retry_interval": 60
}
]
"job_schedule": {
"time_created": int (timestamp)
"time_started": int (timestamp)
"time_ended": int (timestamp)
"status": string ["stop", "scheduled", "running", "aborting", "removed"]
"event": string ["", "stop", "start", "abort", "remove"]
"result": string ["", "success", "fail", "aborted"]
SCHEDULING TIME INFORMATION
}
- Three types of scheduling can be identified:
- date - used for single run jobs
- interval - periodic jobs, providing an interval value
- cron-like jobs
Each type has specific parameters which can be given.
"schedule_date": : datetime isoformat
"schedule_interval" : "continuous", "N weeks" / "N days" / "N hours" / "N minutes" / "N seconds" "schedule_start_date" : datetime isoformat "schedule_end_date" : datetime isoformat
"schedule_year" : 4 digit year "schedule_month" : 1-12 "schedule_day" : 1-31 "schedule_week" : 1-53 "schedule_day_of_week": 0-6 or string mon,tue,wed,thu,fri,sat,sun "schedule_hour" : 0-23 "schedule_minute" : 0-59 "schedule_second" : 0-59 "schedule_start_date" : datetime isoformat "schedule_end_date" : datetime isoformat
example backup freezer_action:
"freezer_action": {
"action" : "backup"
"mode" : "fs"
"src_file" : "/home/tylerdurden/project_mayhem"
"backup_name" : "project_mayhem_backup"
"container" : "my_backup_container"
"max_backup_level" : int
"always_backup_level": int
"restart_always_backup": int
"no_incremental" : bool
"encrypt_pass_file" : private_key_file
"log_file" : "/var/log/freezer.log"
"hostname" : false
"max_cpu_priority" : false
}
example restore freezer_action:
"freezer_action": {
"action": "restore"
"restore-abs-path": "/home/tylerdurden/project_mayhem"
"container" : "my_backup_container"
"backup-name": "project_mayhem_backup"
"restore-from-host": "another_host"
"max_cpu_priority": true
}
example scheduled backup job. job will be executed once at the provided datetime:
"job": {
"job_actions":
[
{
"freezer_action":
{
"action" : "backup",
"mode" : "fs",
"src_file" : "/home/tylerdurden/project_mayhem",
"backup_name" : "project_mayhem_backup",
"container" : "my_backup_container",
}
"exit_status": "fail|success"
"max_retries": int,
"max_retries_interval": secs,
"mandatory": bool
},
{
action
...
},
{
action
...
}
],
"job_schedule":
{
"time_created": 1234,
"time_started": 1234,
"time_ended": 0,
"status": "stop | scheduled | running",
"schedule_date": "2015-06-02T16:20:00",
}
"job_id": "blabla",
"client_id": "blabla",
"user_id": "blabla",
"description": "scheduled one shot",
}
"job": {
"job_actions":
[ ... ],
"job_schedule":
{
"time_created": 1234,
"time_started": 1234,
"time_ended": 0,
"status": "stop",
"event": "start"
"schedule_interval" : "1 day"
"schedule_start_date" : "2015-06-02T16:20:00"
},
"job_id": "blabla",
"client_id": "blabla",
"user_id": "blabla",
"description": "daily backup",
}
multiple scheduling choices allowed:
"job": {
"job_actions":
[ ... ],
"job_schedule":
{
"time_created": 1234,
"time_started": 1234,
"time_ended": 0,
"status": "scheduled"
"schedule_month" : "1-6, 9-12"
"schedule_day" : "mon, wed, fri"
"schedule_hour": "03"
"schedule_minute": "25"
}
"job_id": "blabla",
"client_id": "blabla",
"user_id": "blabla",
"description": "daily backup",
}
Finished job with result:
"job": {
"job_actions": [ ... ],
"job_schedule":
{
"time_created": 1234,
"time_started": 1234,
"time_ended": 4321,
"status": "stop",
"event": "",
"result": "success",
"schedule_time": "2015-06-02T16:20:00"
},
"job_id": "blabla",
"client_id": "blabla",
"user_id": "blabla",
"description": "one shot job",
}
It is possible to define properties that span accross multiple actions This allow not to rewrite values that might be the same in multiple actions. If properties are specificaly set in one action, then the specified value is the one used.
- Example::
- "job": {
- "action_defaults": {
- "log_file": "/tmp/freezer_tmp_log", "container": "my_backup_container"
}, "job_actions":
- [
- {"freezer_action":{
- "action" : "backup", "mode" : "fs", "src_file" : "/home/user1/file", "backup_name" : "user1_backup"}},
- {"freezer_action":{
- "action" : "backup", "mode" : "fs", "src_file" : "/home/user2/file", "backup_name" : "user2_backup"}},
- {"freezer_action":{
- "action" : "backup", "mode" : "fs", "src_file" : "/home/user3/file", "backup_name" : "user2_backup", "log_file" : "/home/user3/specific_log_file" }}
]
"description": "scheduled one shot"
}
- Is Equivalent to::
- "job": {
- "job_actions":
- [
- {"freezer_action":{
- "action" : "backup", "mode" : "fs", "src_file" : "/home/user1/file", "backup_name" : "user1_backup", "log_file": "/tmp/freezer_tmp_log", "container": "my_backup_container"}},
- {"freezer_action":{
- "action" : "backup", "mode" : "fs", "src_file" : "/home/user2/file", "backup_name" : "user2_backup", "log_file": "/tmp/freezer_tmp_log", "container": "my_backup_container"}},
- {"freezer_action":{
- "action" : "backup", "mode" : "fs", "src_file" : "/home/user3/file", "backup_name" : "user2_backup", "log_file" : "/home/user3/specific_log_file", "container": "my_backup_container"}},
]
"description": "scheduled one shot"
}
Actions are stored only to facilitate the assembling of different actions into jobs in the web UI. They are not directly used by the scheduler. They are stored in this structure:
{
"freezer_action": {
"action": string,
"backup_name": string,
....
},
"mandatory": bool,
"max_retries": int,
"max_retries_interval": int
"action_id": string,
"user_id": string
}
A session is a group of jobs which share the same scheduling time. A session is identified by its session_id and has a numeric tag (session_tag) which is incremented each time that a new session is started. The purpose of the session_tag is that of identifying a group of jobs which have been executed together and which therefore represent a snapshot of a distributed system.
When a job is added to a session, the scheduling time of the session is copied into the job data structure, so that any job belonging to the same session will start at the same time.
session =
{
"session_id": string,
"session_tag": int,
"description": string,
"hold_off": int (seconds),
"schedule": { scheduling information, same as jobs },
"jobs": { 'job_id_1': {
"client_id": string,
"status": string,
"result": string
"time_started": int (timestamp),
"time_ended": int (timestamp),
},
'job_id_2': {
"client_id": string,
"status": string,
"result": string
"time_started": int (timestamp),
"time_ended": int (timestamp),
}
}
"time_start": int timestam,
"time_end": int timestam,
"time_started": int (timestamp),
"time_ended": int (timestamp),
"status": string "completed" "running",
"result": string "success" "fail",
"user_id": string
}
When the freezer scheduler running on a node wants to start a session, it sends a POST request to the following endpoint:
POST /v1/sessions/{sessions_id}/action
The body of the request bears the action and parameters
{
"start": {
"job_id": "JOB_ID_HERE",
"current_tag": 22
}
}
Example of a succesfull response:
{
'result': 'success',
'session_tag': 23
}
{
"end": {
"job_id": "JOB_ID_HERE",
"current_tag": 23,
"result": "success|fail"
}
}
PUT /v1/sessions/{sessions_id}/jobs/{job_id} adds the job to the session DELETE /v1/sessions/{sessions_id}/jobs/{job_id} adds the job to the session