Migrate DockerHub Images to GitLab : Script



# Change these for the target image / group.

# Calculate the GitLab image name.

# Pull the image from DockerHub
docker pull $DOCKERHUB_IMAGE

# Tag the image with the GitLab Container Registry path.

# Push the image to the GitLab Container Registry.
docker login registry.gitlab.com -u unused -p $CRED 
docker push $GITLAB_IMAGE

You can generate deploy tokens (R/W) in your project settings. Group level tokens will let this operate across multiple projects in a group.

Sorting S3 Buckets by Size

It can be fairly hard to rank your s3 buckets by size, especially with intelligent tiering on. Here is a concise script to find all bucket sizes in your account using cloudwatch metrics, that will output the top 10 in sorted order.

import boto3
import pandas as pd
from datetime import datetime, timedelta
import logging

# Configure logging
logging.basicConfig(format='%(asctime)s %(levelname)s: %(message)s', level=logging.INFO)

# Connect to CloudWatch
cloudwatch = boto3.client('cloudwatch')

# Connect to S3
s3 = boto3.resource('s3')

# Define a function to get the BucketSizeBytes metric data for a given bucket and storage type
def get_metric_data(bucket, storage_type):
    response = cloudwatch.get_metric_statistics(
            {'Name': 'BucketName', 'Value': bucket},
            {'Name': 'StorageType', 'Value': storage_type}
        StartTime=datetime.utcnow() - timedelta(days=3),
    datapoints = response['Datapoints']
    if datapoints:
        return max([datapoint['Maximum'] for datapoint in datapoints])
        return 0

# Log before pulling the list of bucket names
logging.info("Getting list of bucket names...")

# Get all buckets in the account
buckets = [bucket.name for bucket in s3.buckets.all()]

# Prepare the MetricDataQueries for all the metrics
metric_data_queries = []
for bucket in buckets:
    logging.info(f"Working on bucket: {bucket}...")
    metric_data_queries.append(get_metric_data(bucket, 'StandardStorage'))
    metric_data_queries.append(get_metric_data(bucket, 'IntelligentTieringIAStorage'))
    metric_data_queries.append(get_metric_data(bucket, 'IntelligentTieringFAStorage'))
    metric_data_queries.append(get_metric_data(bucket, 'IntelligentTieringAIAStorage'))

# Parse the MetricData and sum up the bucket sizes
bucket_sizes = {}
for i in range(0, len(metric_data_queries), 4):
    bucket = buckets[i // 4]
    total_size = sum(metric_data_queries[i:i+4])
    bucket_sizes[bucket] = total_size

# Convert the results to a Pandas dataframe and display without truncation
df = pd.DataFrame.from_dict(bucket_sizes, orient='index', columns=['Size (Bytes)'])
df['Size (TBs)'] = df['Size (Bytes)'] / (1024 ** 4)
df = df[['Size (TBs)']].sort_values(by='Size (TBs)', ascending=False).head(10)
pd.set_option('display.max_rows', None)
pd.set_option('display.max_colwidth', None)
pd.set_option('display.width', None)
pd.set_option('display.float_format', '{:.2f}'.format)

Kubectl – View Pods Per Node in Kubernetes

You can use this command to view how many pods are on each node in Kubernetes using just kubect.

kubectl get pods -A -o=custom-columns=NODE:.spec.nodeName | sort | uniq -c | sort -n

In our case, we have a limit of 25 pods per node, so we have daemon sets fail to roll out if nodes already have 25 pods. So, this is helpful.

It can also be helpful when decoming nodes as you track the removal of pods from them.

Querying LDAP From Python the Easy Way

Historically, using LDAP in python could be fairly painful because you had to install python-ldap, which could be hard depending on your environment. E.g. getting that installed in a Jupyter notebook where I work proved impossible without changing the underlying docker image for the notebook.

Most search results will still lead you to python-dap, but now you can and should use python-ldap3 instead. This library is pure-python and does not have any awkward OS dependencies. So, it “just works” and is much lighter.

Here is an example of how to login with a service account and query a user via email.

import ldap3

# Put in params up top.

search_base = 'DC=foo,DC=bar'
search_filter = '(&(mail=john.doe@somecompany.com))'
attrs = ["*"]

server = ldap3.Server(LDAP_URI)
with ldap3.Connection(server, auto_bind=True, user=SERVICE_ACCOUNT, password=SERVICE_ACCOUNT_PASSWORD) as conn:
    conn.search(search_base, search_filter, attributes=attrs)

To run this, you just have to do a quick pip install as shown below. I recommend you use the latest version; but I locked it here just to remind you that locking a version is smart in most python projects. Version drift causes many production issues.

pip install ldap3==2.9.1

Using Presto/Trino CLI w/ TLS & Passwords Enabled

Not much of a post here, just recording a good way to automatically run queries in the presto cli.

./presto-cli-350-executable.jar \
      --server https://your.cluster.dns:443 \
      --catalog hive \
      --schema default \
      --client-request-timeout "10s" \
      --user "john.humphreys" \
      --password \
      --execute "select * from some_db.some_table limit 10"

If you need the password as well, you can pipe it into the command with the yes command in Linux. I also suspect there is an environment variable you can pre-set for that, but I didn’t dig in to double check.