Bug: logger breaks on nodejs Docker lambdas since 2.4.0

[ralbertazzi]

The base image used to build Docker images, public.ecr.aws/lambda/nodejs:20, sets the environment variable TZ=:/etc/localtime. This value is not recognised as valid one and throws an error. I believe this was introduced by this PR that was merged in 2.4.0.

Expected Behavior

Building a Docker Lambda using the logger library without further configuration, i.e. I'd expect the library to print times in UTC when TZ is :/etc/localtime

Current Behavior

RangeError: Invalid time zone specified: :/etc/localtime is raised when logging anything under this condition

Code snippet

import { Logger } from "@aws-lambda-powertools/logger";

const log = new Logger();
log.info("Hello");

Steps to Reproduce

1. Build a Docker Lambda from public.ecr.aws/lambda/nodejs:20
2. Install the logger library
3. Log something

Possible Solution

The library should print times in UTC when TZ is :/etc/localtime

Powertools for AWS Lambda (TypeScript) version

latest

AWS Lambda function runtime

20.x

Packaging format used

npm

Execution logs

RangeError: Invalid time zone specified: :/etc/localtime
    at new DateTimeFormat (<anonymous>)
    at #getDateFormatter (/var/task/node_modules/.pnpm/@aws-lambda-powertools+logger@2.5.0/node_modules/@aws-lambda-powertools/logger/lib/cjs/formatter/LogFormatter.js:85:16)
    at #generateISOTimestampWithOffset (/var/task/node_modules/.pnpm/@aws-lambda-powertools+logger@2.5.0/node_modules/@aws-lambda-powertools/logger/lib/cjs/formatter/LogFormatter.js:103:82)
    at PowertoolsLogFormatter.formatTimestamp (/var/task/node_modules/.pnpm/@aws-lambda-powertools+logger@2.5.0/node_modules/@aws-lambda-powertools/logger/lib/cjs/formatter/LogFormatter.js:53:56)
    at PowertoolsLogFormatter.formatAttributes (/var/task/node_modules/.pnpm/@aws-lambda-powertools+logger@2.5.0/node_modules/@aws-lambda-powertools/logger/lib/cjs/formatter/PowertoolsLogFormatter.js:32:29)
    at Logger.createAndPopulateLogItem (/var/task/node_modules/.pnpm/@aws-lambda-powertools+logger@2.5.0/node_modules/@aws-lambda-powertools/logger/lib/cjs/Logger.js:722:39)
    at Logger.processLogItem (/var/task/node_modules/.pnpm/@aws-lambda-powertools+logger@2.5.0/node_modules/@aws-lambda-powertools/logger/lib/cjs/Logger.js:827:46)
    at Logger.info (/var/task/node_modules/.pnpm/@aws-lambda-powertools+logger@2.5.0/node_modules/@aws-lambda-power

[boring-cyborg]

Thanks for opening your first issue here! We'll come back to you as soon as we can.
In the meantime, check out the #typescript channel on our Powertools for AWS Lambda Discord: Invite link

[dreamorosi]

Hi @ralbertazzi - thank you for opening this issue.

I confirm that I am able to reproduce the issue:

I agree also with the fix proposed, if the value is :/etc/localtime we should default to :UTC. The fix should be applied in this function, which should be modified to do something like this:

public formatTimestamp(now: Date): string {
  const defaultTimezone = 'UTC';

  /**
   * If a specific timezone is configured and it's not the default `UTC`,
   * format the timestamp with the appropriate timezone offset.
   **/
- const configuredTimezone = this.envVarsService?.getTimezone();
+ const envVartimezone = this.envVarsService?.getTimezone();
+ const configuredTimezone =
+   envVartimezone === ':/etc/localtime' ? defaultTimezone : envVartimezone;
  if (configuredTimezone && !configuredTimezone.includes(defaultTimezone))
    return this.#generateISOTimestampWithOffset(now, configuredTimezone);

  return now.toISOString();
}

Additionally, we should also add a test case at the bottom of this block here to test for this case, similar to:

it('uses the default UTC timezone when the TZ env variable is set to `:/etc/localtime`', () => {
  // Prepare
  process.env.TZ = ':/etc/localtime';
  const formatter = new PowertoolsLogFormatter({
    envVarsService: new EnvironmentVariablesService(),
  });

  // Act
  const timestamp = formatter.formatTimestamp(new Date());

  // Assess
  expect(timestamp).toEqual('2016-06-20T12:08:10.000Z');
});

I'll mark the issue as help-wanted and good-first-issue to signify that contributions are very welcome.

If nobody picks it up, we'll work on it and include the fix in the next release (~next Wednesday).

In the meantime, as a stopgap solution you can set the TZ environment variable to :UTC to continue using this version of Logger.

In parallel, I'll also consult the AWS Lambda team to see if the container image should align its behavior with the managed runtime & set the value to :UTC.