使用SQL查询检测两个或多个连续行是否具有相同条目

Question

I have an "alert" table in my database, that stores my web apps status codes. These status codes are obtained from a task scheduler, written in python, that periodically queries the front and backend api, and stores the resulting status code returned in the response into their respective columns: frontend_status and backend_status .

I'm trying to do a simple (at least the idea sounds simple) sql query to check if two or more consecutive entries in one of these columns is anything other than 200 (meaning something could be wrong). If it is, I want to use a boolean to take action. I am not sure if I am missing something obvious but some help would be great.

I created a table in postgres, that looks something like this:

-- Table: public.alert_001

-- DROP TABLE public.alert_001;

CREATE TABLE public.alert_001
(
    data_id integer NOT NULL DEFAULT nextval('alert_001_data_id_seq'::regclass),
    when_captured timestamp without time zone NOT NULL,
    frontend_status real,
    backend_status real,
    CONSTRAINT alert_001_pkey PRIMARY KEY (data_id)
)
WITH (
    OIDS = FALSE
)
TABLESPACE pg_default;

ALTER TABLE public.alert_001
    OWNER to postgres;

-- Index: alert_001_when_index

-- DROP INDEX public.alert_001_when_index;

CREATE INDEX alert_001_when_index
    ON public.alert_001 USING btree
    (when_captured)
    TABLESPACE pg_default;

I then managed to write a query that counts the number of entries that are a certain value in the last hour. If the counted entries are more than 2 for example, it will return a boolean true. However, although this works when two entries are picked up in the last hour, they are not necessarily consecutive entries. So my sql query looks something like:

SELECT (SELECT COUNT(*) AS "Count" 
FROM alert_001 WHERE when_captured > NOW() - '1 hour'::INTERVAL AND backend_status != 200) > 2

Could anyone point me into the right direction to catch three or more consecutive entries that are not of status code 200?

Here is an example of the entries in the alert_001 table:

when_captured,frontend_status, backend_status
'2018-02-02 14:55:19.63941','200','200'
'2018-02-02 14:54:19.636386','200','503'
'2018-02-02 14:53:19.636055','200','503'
'2018-02-02 14:52:19.631958','200','503'
'2018-02-02 14:51:19.62166','200','200'
'2018-02-02 14:50:19.621363','200','200'
'2018-02-02 14:49:19.612434','200','200'
'2018-02-02 14:48:19.611919','200','200'
'2018-02-02 14:47:19.610065','200','200'
'2018-02-02 14:46:19.607846','200','200'

And as you can see there are three consecutive entries with a 503 backend_status code, which I want to catch. So, the expected output of the correct sql query could be True if more than three are found or False otherwise?

Answer 1

You should use window functions for this:

SELECT EXISTS (
   SELECT 1 FROM
      (SELECT data_id,
              backend_status status,
              lag(backend_status) OVER w status_1,
              lag(backend_status, 2) OVER w status_2
         FROM alert_001
         WHERE when_captured > current_timestamp - INTERVAL '1 hour'
         WINDOW w AS (ORDER BY when_captured)
      ) last_three
   WHERE status <> 200
     AND status_1 <> 200
     AND status_2 <> 200
);