如何使用 CLI 删除 AWS S3中版本化的 bucket？

One way to do it is iterate through the versions and delete them. A bit tricky on the CLI, but as you mentioned Java, that would be more straightforward:

AmazonS3Client s3 = new AmazonS3Client();
String bucketName = "deleteversions-"+UUID.randomUUID();


//Creates Bucket
s3.createBucket(bucketName);


//Enable Versioning
BucketVersioningConfiguration configuration = new BucketVersioningConfiguration(ENABLED);
s3.setBucketVersioningConfiguration(new SetBucketVersioningConfigurationRequest(bucketName, configuration ));


//Puts versions
s3.putObject(bucketName, "some-key",new ByteArrayInputStream("some-bytes".getBytes()), null);
s3.putObject(bucketName, "some-key",new ByteArrayInputStream("other-bytes".getBytes()), null);


//Removes all versions
for ( S3VersionSummary version : S3Versions.inBucket(s3, bucketName) ) {
String key = version.getKey();
String versionId = version.getVersionId();
s3.deleteVersion(bucketName, key, versionId);
}


//Removes the bucket
s3.deleteBucket(bucketName);
System.out.println("Done!");

You can also batch delete calls for efficiency if needed.

You can delete all the objects in the versioned s3 bucket. But I don't know how to delete specific objects.

$ aws s3api delete-objects \
--bucket <value> \
--delete "$(aws s3api list-object-versions \
--bucket <value> | \
jq '{Objects: [.Versions[] | {Key:.Key, VersionId : .VersionId}], Quiet: false}')"

Alternatively without jq:

$ aws s3api delete-objects \
--bucket ${bucket_name} \
--delete "$(aws s3api list-object-versions \
--bucket "${bucket_name}" \
--output=json \
--query='{Objects: Versions[].{Key:Key,VersionId:VersionId}}')"

I ran into the same limitation of the AWS CLI. I found the easiest solution to be to use Python and boto3:

#!/usr/bin/env python


BUCKET = 'your-bucket-here'


import boto3


s3 = boto3.resource('s3')
bucket = s3.Bucket(BUCKET)
bucket.object_versions.delete()


# if you want to delete the now-empty bucket as well, uncomment this line:
#bucket.delete()

A previous version of this answer used boto but that solution had performance issues with large numbers of keys as Chuckles pointed out.

1. For deleting specify object(s), using jq filter.
2. You may need cleanup the 'DeleteMarkers' not just 'Versions'.
3. Using $() instead of ``, you may embed variables for bucket-name and key-value.

aws s3api delete-objects --bucket bucket-name --delete "$(aws s3api list-object-versions --bucket bucket-name | jq -M '{Objects: [.["Versions","DeleteMarkers"][]|select(.Key == "key-value")| {Key:.Key, VersionId : .VersionId}], Quiet: false}')"

Here is a one liner you can just cut and paste into the command line to delete all versions and delete markers (it requires aws tools, replace yourbucket-name-backup with your bucket name)

echo '#!/bin/bash' > deleteBucketScript.sh \
&& aws --output text s3api list-object-versions --bucket $BUCKET_TO_PERGE \
| grep -E "^VERSIONS" |\
awk '{print "aws s3api delete-object --bucket $BUCKET_TO_PERGE --key "$4" --version-id "$8";"}' >> \
deleteBucketScript.sh && . deleteBucketScript.sh; rm -f deleteBucketScript.sh; echo '#!/bin/bash' > \
deleteBucketScript.sh && aws --output text s3api list-object-versions --bucket $BUCKET_TO_PERGE \
| grep -E "^DELETEMARKERS" | grep -v "null" \
| awk '{print "aws s3api delete-object --bucket $BUCKET_TO_PERGE --key "$3" --version-id "$5";"}' >> \
deleteBucketScript.sh && . deleteBucketScript.sh; rm -f deleteBucketScript.sh;

then you could use:

aws s3 rb s3://bucket-name --force

I ran into issues with Abe's solution as the list_buckets generator is used to create a massive list called all_keys and I spent an hour without it ever completing. This tweak seems to work better for me, I had close to a million objects in my bucket and counting!

import boto


s3 = boto.connect_s3()
bucket = s3.get_bucket("your-bucket-name-here")


chunk_counter = 0 #this is simply a nice to have
keys = []
for key in bucket.list_versions():
keys.append(key)
if len(keys) > 1000:
bucket.delete_keys(keys)
chunk_counter += 1
keys = []
print("Another 1000 done.... {n} chunks so far".format(n=chunk_counter))


#bucket.delete() #as per usual uncomment if you're sure!

Hopefully this helps anyone else encountering this S3 nightmare!

Using boto3 it's even easier than with the proposed boto solution to delete all object versions in an S3 bucket:

#!/usr/bin/env python
import boto3


s3 = boto3.resource('s3')
bucket = s3.Bucket('your-bucket-name')
bucket.object_versions.all().delete()

Works fine also for very large amounts of object versions, although it might take some time in that case.

I found the other answers either incomplete or requiring external dependencies to be installed (like boto), so here is one that is inspired by those but goes a little deeper.

As documented in Working with Delete Markers, before a versioned bucket can be removed, all its versions must be completely deleted, which is a 2-step process:

1. "delete" all version objects in the bucket, which marks them as deleted but does not actually delete them
2. complete the deletion by deleting all the deletion marker objects

Here is the pure CLI solution that worked for me (inspired by the other answers):

#!/usr/bin/env bash


bucket_name=...


del_s3_bucket_obj()
{
local bucket_name=$1
local obj_type=$2
local query="{Objects: $obj_type[].{Key:Key,VersionId:VersionId}}"
local s3_objects=$(aws s3api list-object-versions --bucket ${bucket_name} --output=json --query="$query")
if ! (echo $s3_objects | grep -q '"Objects": null'); then
aws s3api delete-objects --bucket "${bucket_name}" --delete "$s3_objects"
fi
}


del_s3_bucket_obj ${bucket_name} 'Versions'
del_s3_bucket_obj ${bucket_name} 'DeleteMarkers'

Once this is done, the following will work:

aws s3 rb "s3://${bucket_name}"

Not sure how it will fare with 1000+ objects though, if anyone can report that would be awesome.

By far the easiest method I've found is to use this CLI tool, s3wipe. It's provided as a docker container so you can use it like so:

$ docker run -it --rm slmingol/s3wipe --help
usage: s3wipe [-h] --path PATH [--id ID] [--key KEY] [--dryrun] [--quiet]
[--batchsize BATCHSIZE] [--maxqueue MAXQUEUE]
[--maxthreads MAXTHREADS] [--delbucket] [--region REGION]


Recursively delete all keys in an S3 path


optional arguments:
-h, --help               show this help message and exit
--path PATH              S3 path to delete (e.g. s3://bucket/path)
--id ID                  Your AWS access key ID
--key KEY                Your AWS secret access key
--dryrun                 Don't delete. Print what we would have deleted
--quiet                  Suprress all non-error output
--batchsize BATCHSIZE    # of keys to batch delete (default 100)
--maxqueue MAXQUEUE      Max size of deletion queue (default 10k)
--maxthreads MAXTHREADS  Max number of threads (default 100)
--delbucket              If S3 path is a bucket path, delete the bucket also
--region REGION          Region of target S3 bucket. Default vaue `us-
east-1`

Example

Here's an example where I'm deleting all the versioned objects in a bucket and then deleting the bucket:

$ docker run -it --rm slmingol/s3wipe \
--id $(aws configure get default.aws_access_key_id) \
--key $(aws configure get default.aws_secret_access_key) \
--path s3://bw-tf-backends-aws-example-logs \
--delbucket
[2019-02-20@03:39:16] INFO: Deleting from bucket: bw-tf-backends-aws-example-logs, path: None
[2019-02-20@03:39:16] INFO: Getting subdirs to feed to list threads
[2019-02-20@03:39:18] INFO: Done deleting keys
[2019-02-20@03:39:18] INFO: Bucket is empty.  Attempting to remove bucket

How it works

There's a bit to unpack here but the above is doing the following:

• docker run -it --rm mikelorant/s3wipe - runs s3wipe container interactively and deletes it after each execution
• --id & --key - passing our access key and access id in
• aws configure get default.aws_access_key_id - retrieves our key id
• aws configure get default.aws_secret_access_key - retrieves our key secret
• --path s3://bw-tf-backends-aws-example-logs - bucket that we want to delete
• --delbucket - deletes bucket once emptied

References

• https://github.com/slmingol/s3wipe
• Is there a way to export an AWS CLI Profile to Environment Variables?
• https://cloud.docker.com/u/slmingol/repository/docker/slmingol/s3wipe

https://gist.github.com/wknapik/191619bfa650b8572115cd07197f3baf

#!/usr/bin/env bash


set -eEo pipefail
shopt -s inherit_errexit >/dev/null 2>&1 || true


if [[ ! "$#" -eq 2 || "$1" != --bucket ]]; then
echo -e "USAGE: $(basename "$0") --bucket <bucket>"
exit 2
fi


# $@ := bucket_name
empty_bucket() {
local -r bucket="${1:?}"
for object_type in Versions DeleteMarkers; do
local opt=() next_token=""
while [[ "$next_token" != null ]]; do
page="$(aws s3api list-object-versions --bucket "$bucket" --output json --max-items 1000 "${opt[@]}" \
--query="[{Objects: ${object_type}[].{Key:Key, VersionId:VersionId}}, NextToken]")"
objects="$(jq -r '.[0]' <<<"$page")"
next_token="$(jq -r '.[1]' <<<"$page")"
case "$(jq -r .Objects <<<"$objects")" in
'[]'|null) break;;
*) opt=(--starting-token "$next_token")
aws s3api delete-objects --bucket "$bucket" --delete "$objects";;
esac
done
done
}


empty_bucket "${2#s3://}"

E.g. empty_bucket.sh --bucket foo

This will delete all object versions and delete markers in a bucket in batches of 1000. Afterwards, the bucket can be deleted with aws s3 rb s3://foo.

Requires bash, awscli and jq.

This bash script found here: https://gist.github.com/weavenet/f40b09847ac17dd99d16

worked as is for me.

I saved script as: delete_all_versions.sh and then simply ran:

./delete_all_versions.sh my_foobar_bucket

and that worked without a flaw.

Did not need python or boto or anything.

This works for me. Maybe running later versions of something and above > 1000 items. been running a couple of million files now. However its still not finished after half a day and no means to validate in AWS GUI =/

# Set bucket name to clearout
BUCKET = 'bucket-to-clear'


import boto3
s3 = boto3.resource('s3')
bucket = s3.Bucket(BUCKET)


max_len         = 1000      # max 1000 items at one req
chunk_counter   = 0         # just to keep track
keys            = []        # collect to delete


# clear files
def clearout():
global bucket
global chunk_counter
global keys
result = bucket.delete_objects(Delete=dict(Objects=keys))


if result["ResponseMetadata"]["HTTPStatusCode"] != 200:
print("Issue with response")
print(result)


chunk_counter += 1
keys = []
print(". {n} chunks so far".format(n=chunk_counter))
return


# start
for key in bucket.object_versions.all():
item = {'Key': key.object_key, 'VersionId': key.id}
keys.append(item)
if len(keys) >= max_len:
clearout()


# make sure last files are cleared as well
if len(keys) > 0:
clearout()


print("")
print("Done, {n} items deleted".format(n=chunk_counter*max_len))
#bucket.delete() #as per usual uncomment if you're sure!

For those using multiple profiles via ~/.aws/config

import boto3


PROFILE = "my_profile"
BUCKET = "my_bucket"


session = boto3.Session(profile_name = PROFILE)
s3 = session.resource('s3')
bucket = s3.Bucket(BUCKET)
bucket.object_versions.delete()

You can do this from the AWS Console using Lifecycle Rules.

Open the bucket in question. Click the Management tab at the top. Make sure the Lifecycle Sub Tab is selected. Click + Add lifecycle rule

On Step 1 (Name and scope) enter a rule name (e.g. removeall) Click Next to Step 2 (Transitions) Leave this as is and click Next.

You are now on the 3. Expiration step. Check the checkboxes for both Current Version and Previous Versions. Click the checkbox for "Expire current version of object" and enter the number 1 for "After _____ days from object creation Click the checkbox for "Permanently delete previous versions" and enter the number 1 for "After _____ days from becoming a previous version"

click the checkbox for "Clean up incomplete multipart uploads" and enter the number 1 for "After ____ days from start of upload" Click Next Review what you just did.
Click Save

Come back in a day and see how it is doing.

This two bash lines are enough for me to enable the bucket deletion !

1: Delete objects aws s3api delete-objects --bucket ${buckettoempty} --delete "$(aws s3api list-object-versions --bucket ${buckettoempty} --query='{Objects: Versions[].{Key:Key,VersionId:VersionId}}')"

2: Delete markers aws s3api delete-objects --bucket ${buckettoempty} --delete "$(aws s3api list-object-versions --bucket ${buckettoempty} --query='{Objects: DeleteMarkers[].{Key:Key,VersionId:VersionId}}')"

To add to python solutions provided here: if you are getting boto.exception.S3ResponseError: S3ResponseError: 400 Bad Request error, try creating ~/.boto file with the following data:

[Credentials]
aws_access_key_id = aws_access_key_id
aws_secret_access_key = aws_secret_access_key
[s3]
host=s3.eu-central-1.amazonaws.com
aws_access_key_id = aws_access_key_id
aws_secret_access_key = aws_secret_access_key

Helped me to delete bucket in Frankfurt region.

Original answer: https://stackoverflow.com/a/41200567/2586441

Even though technically it's not AWS CLI, I'd recommend using AWS Tools for Powershell for this task. Then you can use the simple command as below:

Remove-S3Bucket -BucketName {bucket-name} -DeleteBucketContent -Force -Region {region}

As stated in the documentation, DeleteBucketContent flag does the following:

"If set, all remaining objects and/or object versions in the bucket are deleted proir (sic) to the bucket itself being deleted"

Reference: https://docs.aws.amazon.com/powershell/latest/reference/items/Remove-S3Bucket.html

If you have to delete/empty large S3 buckets, it becomes quite inefficient (and expensive) to delete every single object and version. It's often more convenient to let AWS expire all objects and versions.

aws s3api put-bucket-lifecycle-configuration \
--lifecycle-configuration '{"Rules":[{
"ID":"empty-bucket",
"Status":"Enabled",
"Prefix":"",
"Expiration":{"Days":1},
"NoncurrentVersionExpiration":{"NoncurrentDays":1}
}]}' \
--bucket YOUR-BUCKET

Then you just have to wait 1 day and the bucket can be deleted with:

aws s3api delete-bucket --bucket YOUR-BUCKET

If you want pure CLI approach (with jq):

aws s3api list-object-versions \
--bucket $bucket \
--region $region \
--query "Versions[].Key"  \
--output json | jq 'unique' | jq -r '.[]' | while read key; do
echo "deleting versions of $key"
aws s3api list-object-versions \
--bucket $bucket \
--region $region \
--prefix $key \
--query "Versions[].VersionId"  \
--output json | jq 'unique' | jq -r '.[]' | while read version; do
echo "deleting $version"
aws s3api delete-object \
--bucket $bucket \
--key $key \
--version-id $version \
--region $region
done
done

Looks like as of now, there is an Empty button in the AWS S3 console.

Just select your bucket and click on it. It will ask you to confirm your decision by typing permanently delete Note, this will not delete the bucket itself.

You can use aws-cli to delete s3 bucket

aws s3 rb s3://your-bucket-name

If aws cli is not installed in your computer you can your following commands: For Linux or ubuntu:

sudo apt-get install aws-cli

Then check it is installed or not by:

aws --version

Now configure it by providing aws-access-credentials

aws configure

Then give the access key and secret access key and your region

Simple bash loop I've found and implemented for N buckets:

for b in $(ListOfBuckets); do \
echo "Emptying $b"; \
aws s3api delete-objects --bucket $b --delete "$(aws s3api list-object-versions --bucket $b --output=json --query='{Objects: *[].{Key:Key,VersionId:VersionId}}')"; \
done

I improved the boto3 answer with Python3 and argv.

1. Save the following script as something like s3_rm.py.

#!/usr/bin/env python3
import sys
import boto3


def main():
args = sys.argv[1:]
if (len(args) < 1):
print("Usage: {} s3_bucket_name".format(sys.argv[0]))
exit()


s3 = boto3.resource('s3')
bucket = s3.Bucket(args[0])
bucket.object_versions.delete()


# if you want to delete the now-empty bucket as well, uncomment this line:
#bucket.delete()


if __name__ == "__main__":
main()

2. Add chmod +x s3_rm.py.
3. Run the function like ./s3_rm.py my_bucket_name.

If you use AWS SDK for JavaScript S3 Client for Node.js (@aws-sdk/client-s3), you can use following code:

const { S3Client, ListObjectsCommand } = require('@aws-sdk/client-s3')




const endpoint = 'YOUR_END_POINT'
const region = 'YOUR_REGION'


// Create an Amazon S3 service client object.
const s3Client = new S3Client({ region, endpoint })


const deleteEverythingInBucket = async bucketName => {
console.log('Deleting all object in the bucket')


const bucketParams = {
Bucket: bucketName
}


try {
const command = new ListObjectsCommand(bucketParams)
const data = await s3Client.send(command)
console.log('Bucket Data', JSON.stringify(data))
if (data?.Contents?.length > 0) {
console.log('Removing objects in the bucket', data.Contents.length)
for (const object of data.Contents) {
console.log('Removing object', object)
if (object.Key) {
try {
await deleteFromS3({
Bucket: bucketName,
Key: object.Key
})
} catch (err) {
console.log('Error on object delete', err)
}
}
}
}
} catch (err) {
console.log('Error creating presigned URL', err)
}
}

For my case, I wanted to be sure that all objects for specific prefixes would be deleted. So, we generate a list of all objects for each prefix, divide it by 1k records (AWS limitation), and delete them.

Please note that AWS CLI and jq must be installed and configured.

A text file with prefixes that we want to delete was created (in the example below prefixes.txt).

The format is:

prefix1
prefix2

And this is a shell script (also please change the BUCKET_NAME with the real name):

#!/bin/sh


BUCKET="BUCKET_NAME"
PREFIXES_FILE="prefixes.txt"


if [ -f "$PREFIXES_FILE" ]; then
while read -r current_prefix
do
printf '***** PREFIX %s *****\n' "$current_prefix"


OLD_OBJECTS_FILE="$current_prefix-all.json"


if [ -f "$OLD_OBJECTS_FILE" ]; then
printf 'Deleted %s...\n' "$OLD_OBJECTS_FILE"


rm "$OLD_OBJECTS_FILE"
fi


cmd="aws s3api list-object-versions --bucket \"$BUCKET\" --prefix \"$current_prefix/\" --query \"[Versions,DeleteMarkers][].{Key: Key, VersionId: VersionId}\"  >> $OLD_OBJECTS_FILE"
echo "$cmd"
eval "$cmd"


no_of_obj=$(cat "$OLD_OBJECTS_FILE" | jq 'length')
i=0
page=0


#Get old version Objects
echo "Objects versions count: $no_of_obj"


while [ $i -lt "$no_of_obj" ]
do
next=$((i+999))
old_versions=$(cat "$OLD_OBJECTS_FILE" | jq '.[] | {Key,VersionId}' | jq -s '.' | jq .[$i:$next])
paged_file_name="$current_prefix-page-$page.json"
cat << EOF > "$paged_file_name"
{"Objects":$old_versions, "Quiet":true}
EOF
echo "Deleting records from $i - $next"
cmd="aws s3api delete-objects --bucket \"$BUCKET\" --delete file://$paged_file_name"
echo "$cmd"
eval "$cmd"


i=$((i+1000))
page=$((page+1))
done


done < "$PREFIXES_FILE"
else
echo "$PREFIXES_FILE does not exist."
fi

If you want just to check the list of objects and don't delete them immediately - please comment/remove the last eval "$cmd".

In the same vein as https://stackoverflow.com/a/63613510/805031 ... this is what I use to clean up accounts before closing them:

# If the data is too large, apply LCP to remove all objects within a day


# Create lifecycle-expire.json with the LCP required to purge all objects
# Based on instructions from: https://aws.amazon.com/premiumsupport/knowledge-center/s3-empty-bucket-lifecycle-rule/
cat << JSON > lifecycle-expire.json
{
"Rules": [
{
"ID": "remove-all-objects-asap",
"Filter": {
"Prefix": ""
},
"Status": "Enabled",
"Expiration": {
"Days": 1
},
"NoncurrentVersionExpiration": {
"NoncurrentDays": 1
},
"AbortIncompleteMultipartUpload": {
"DaysAfterInitiation": 1
}
},
{
"ID": "remove-expired-delete-markers",
"Filter": {
"Prefix": ""
},
"Status": "Enabled",
"Expiration": {
"ExpiredObjectDeleteMarker": true
}
}
]
}
JSON


# Apply to ALL buckets
aws s3 ls | cut -d" " -f 3 | xargs -I{} aws s3api put-bucket-lifecycle-configuration --bucket {} --lifecycle-configuration file://lifecycle-expire.json


# Apply to a single bucket; replace $BUCKET_NAME
aws s3api put-bucket-lifecycle-configuration --bucket $BUCKET_NAME --lifecycle-configuration file://lifecycle-expire.json

...then a day later you can come back and delete the buckets using something like:

# To force empty/delete all buckets
aws s3 ls | cut -d" " -f 3 | xargs -I{} aws s3 rb s3://{} --force


# To remove only empty buckets
aws s3 ls | cut -d" " -f 3 | xargs -I{} aws s3 rb s3://{}


# To force empty/delete a single bucket; replace $BUCKET_NAME
aws s3 rb s3://$BUCKET_NAME --force

It saves a lot of time and money so worth doing when you have many TBs to delete.

I needed to delete older object versions but keep the current version in the bucket. Code uses iterators, works on buckets of any size with any number of objects.

import boto3
from itertools import islice


bucket = boto3.resource('s3').Bucket('bucket_name'))
all_versions = bucket.object_versions.all()
stale_versions = iter(filter(lambda x: not x.is_latest, all_versions))
pages = iter(lambda: tuple(islice(stale_versions, 1000)), ())
for page in pages:
bucket.delete_objects(
Delete={
'Objects': [{
'Key': item.key,
'VersionId': item.version_id
} for item in page]
})