请仔细考虑重新格式化查询，因为在某些情况下，当sql解析本身由于诸如空格之类的简单问题而失败时，impala会与segv崩溃。如果您运行的是cloudera，则会在 /run/cloudera-scm-agent/process 在运行查询的节点上。
我们通过小心处理sql格式（这也是一个很好的实践，因为它使查询错误更容易发现）解决了这些问题。

SELECT
    COUNT(p.id)                                                     AS tweet_count,
    p.author                                                        AS author,
    p.profile_image_url                                             AS profile_image_url,
    p.screen_name                                                   AS screen_name,
    concat_ws('/', MIN(p.postday), MIN(p.postmonth), MIN(p.postyear) ) AS creation_date,
    p.message                                                       AS MESSAGE,
    af.followerid                                                   AS follower
FROM
    post p
INNER JOIN
    author_follower af
ON
    af.id LIKE IF(p.author = NULL, '', concat(p.author, '%'))
WHERE
    p.hashtaglist LIKE 'hashtagtobeused'
GROUP BY
    author,
    profile_image_url,
    screen_name,
    MESSAGE,
    follower
ORDER BY
    CAST(MIN(postyear) AS INT),
    CAST(MIN(postmonth) AS INT),
    CAST(MIN(postday) AS INT),
    CAST(MIN(posthour) AS INT) ASC;

（顺便说一句，我使用dbvisualizer验证并重新格式化查询语法——这是一个很好的工具）