lilo-cli

lilo

lilo is a CLI to download and tail Google Cloud Platform logs to a SQLite database.

# Install
npm install -g bun
npm install -g lilo-cli

# Note: `gcloud` also needs to be installed and signed in. 

# Usage
lilo \
    --resource-names "projects/YOUR_GCP_PROJECT_ID_HERE" \
    --filter "timestamp > 2023-06-01" \
    --db ./db.sqlite \
    --watch 2000

See lilo help for detailed flag docs.

• Log JSON values from your application and ask questions later with SQL queries.

  • Avoid creating a data schema for ad-hoc events/data when developing prototypes.
  • GCP Logging supports JSON lines - your application can log JSON lines to stdout.

• Use fast & familiar local tools.

  • Use a native GUI such as TablePlus for viewing tables.
  • Use SQLites SQL dialect instead of BigQuery.
  • SQLite indexes on expressions can index a JSON path to speed up queries.
    • Like this: create index i_01 on logs(jsonPayload ->> "$.msg")
  • Use the SQLite CLI to query and pipe output JSON to Excel or other tools to create reports.
  • Any programming language that supports SQLite.
  • Fast interactivity for small to medium datasets.

• Avoid having to store large amounts of data forever in the cloud.

  • Set your Logging retention to one month, and use lilo to archive your data to a local SQLite DB.

• Query the db.sqlite on the server to respond to events.

  • Use polling SQL queries to observe events.

• Understand what's happening on your GCP account/audit logs.

• Convert the SQLite file to DuckDB for fast analytical queries.

Name origin: SQLite GCP Logs

Example table schema and rows in TablePlus

FAQ

How long does it take to do a full download?

• It depends on how many logs you have, but by default GCP accounts are limited to 60 HTTP read calls per minute, and each of those has 1000 entries.
• lilo begins downloading from the last log entry it downloaded.
  • You can start and stop the CLI at any time.
  • lilo only downloads the log entries since the last run.

Why Bun?

• It has SQLite built in, so SQLite will not need to be compiled when installing lilo via NPM.
• I wanted to test it out.

Alternatives & Notes

• Set up a log sink to BigQuery, then pipe a BigQuery query result to Google Sheets.

• Copy your existing logs to a Storage Bucket

  • Use for log entries that were written before you set up a streaming log sink.
  • https://cloud.google.com/logging/docs/routing/copy-logs

• https://cloud.google.com/logging/docs/samples/logging-tail-log-entries

• https://cloud.google.com/logging/docs/reference/v2/rest/v2/entries/tail

  • The tail endpoint does not work with JSON, only GRPC.
  • Bun does not currently support GRPC, so lilo polls entries:list

• gcloud alpha logging tail

• https://til.simonwillison.net/cloudrun/tailing-cloud-run-request-logs