This script looks for the most recent snapshot for a given DBInstanceIdentifier
and restores it to a new instance in the same or different VPC.
It is designed to only have one restored instance at once (it will spin up the new one then delete the old). Once the old istance is deleted, it changes the DNS record to point to the new instance. It should only take a 15 or so minutes to run (tested on a 600GB m4.2xlarge).
- You want to have a staging database with similar or identical data to production to test your applications against
- You want to test your production snapshots
- You really like creating new RDS instances
- Probably others I haven't though of
To run this in kubernetes:
- Fill out the container arguments in
cronjob.yml
to suit your environment - especially if you want to build your own image - Create a kubernetes secret called
rds-snapshot-restore
with your AWS Credentials in it and add it to your cluster kubectl apply -f cronjob.yml
Please open an issue if you find a bug or have a feature request (or if you think I did something stupid).
The following targets an RDS instance named mydata
and will restore it to an instance called staging-mydata-[some unix timestamp]
, create a CNAME called mydata.staging.example.com
in a Route53 zone with staging
in the name (specified by -m
). Since -V
and -S
are specified, the new instance will be in a different VPC. The VPC used matches the Name: staging
key/value tag and the database
security group is used. Since --noop
is used, no action is done.
./snapshot_restore.py -p staging -t mydata -u staging-db-subnet-group -S database -c mydata -m staging -D staging.example.com --noop -V staging
Another scenario would be to restore the snapshot to an instance in the same VPC but with a prefix staging
on its name - staging-mydata.example.com
:
./snapshot_restore.py -p staging -t mydata -c mydata -z Z9999999999
This script assumes:
- You have a running RDS instance with at least one snapshot
- You use Route53 hosted zones
- Your RDS instances are in VPCs
usage: snapshot_restore.py [-h] -t TARGET_INSTANCE [-r AWS_REGION]
[-i INSTANCE_CLASS] [-p PREFIX]
[-u SUBNET_GROUP_NAME]
[-S SECURITY_GROUP_NAMES [SECURITY_GROUP_NAMES ...]]
[-V VPC_TAG_NAME] -D DNS_SUFFIX -c CNAME_NAME
[-m ZONE_MATCH_STRING] [-z ZONE_ID]
[-s SNAPSHOT_TYPE] [-n]
Restore the most recent snapshot of a given RDS instance to a new instance
optional arguments:
-h, --help show this help message and exit
-t TARGET_INSTANCE, --target TARGET_INSTANCE
Name of the RDS instance to restore the snapshot from
- this is the DBInstanceIdentifier of the target
instance
-r AWS_REGION, --region AWS_REGION
AWS Region to use. Currently can only target and
restore to the same region. Defaults to us-west-2.
-i INSTANCE_CLASS, --instance-class INSTANCE_CLASS
Instance class to use for the new instance. Defaults
to db.t2.medium
-p PREFIX, --prefix PREFIX
Prefix for the new instance DBInstanceIdentifier
-u SUBNET_GROUP_NAME, --subnet-name SUBNET_GROUP_NAME
Name of the database subnet group to use for the new
instance. Defaults to the subnet group of the target
instance if not specified
-S SECURITY_GROUP_NAMES [SECURITY_GROUP_NAMES ...], --sec-group-names SECURITY_GROUP_NAMES [SECURITY_GROUP_NAMES ...]
Names of the VPC security group to use for the new
instance. Defaults to the security group of the target
instance if not specified. Specify multiple separated
by a space. Must also specify --vpc-tag-name
-V VPC_TAG_NAME, --vpc-tag-name VPC_TAG_NAME
VPC "Name" tag value. Required when using --sec-group-names
-D DNS_SUFFIX, --dns-suffix DNS_SUFFIX
DNS Suffix for your private Route53 zone
-c CNAME_NAME, --cname-name CNAME_NAME
Name of the CNAME to create for the new instance
-m ZONE_MATCH_STRING, --match-zone ZONE_MATCH_STRING
String to match a Route53 Hosted Zone name on to
determine the zone ID. Useful if you rebuild private
DNS zones often. Overrides --zone-id if specified.
-z ZONE_ID, --zone-id ZONE_ID
Route53 Zone ID to use for the new instance CNAME.
Defaults to the zone id of the target instance. Is
overridden by --match-zone.
-s SNAPSHOT_TYPE, --snapshot-type SNAPSHOT_TYPE
Snapshot type to search filter on. Defaults to 'automated'.
-n, --noop Enable NOOP mode - will not perform any restore tasks.
Please fork and PR to master if you have something to add