If the body throws an exception, mark the build as a failure, but nonetheless continue to execute the workflow from the statement following the catchError step. This is only necessary when using certain post-build actions (notifiers) originally defined for freestyle projects which pay attention to the result of the ongoing build.

node {
    catchError {
        sh 'might fail'
    }
    step([$class: 'Mailer', recipients: 'admin@somewhere'])
}

If the shell step fails, the flow build’s status will be set to failed, so that the subsequent mail step will see that this build is failed. In the case of the mail sender, this means that it will send mail. (It may also send mail if this build succeeded but previous ones failed, and so on.) Even in that case, this step can be replaced by the following idiom:

node {
    try {
        sh 'might fail'
    } catch (err) {
        echo "Caught: ${err}"
        currentBuild.result = 'FAILURE'
    }
    step([$class: 'Mailer', recipients: 'admin@somewhere'])
}

For all other cases, use plain try-catch(-finally) blocks:

node {
    sh './set-up.sh'
    try {
        sh 'might fail'
        echo 'Succeeded!'
    } catch (err) {
        echo "Failed: ${err}"
    } finally {
        sh './tear-down.sh'
    }
    echo 'Printed whether above succeeded or failed.'
}
// …and the workflow as a whole succeeds

See this document for background.