Don't fail if there were downtimes during the reporting period

[bluthg]

I'm a bit puzzled....

• What's the issue?
• Where's the test showing the issue?
• Do we really need this monstrosity of a function? I'm pretty sure there's a much nices (and less performance hungry!) way to get around this...

[cla-bot[bot]]

Thank you for your pull request. Before we can look at it, you'll need to sign a Contributor License Agreement (CLA).

Please follow instructions at https://icinga.com/company/contributor-agreement to sign the CLA.

After that, please reply here with a comment and we'll verify.

Contributors that have not signed yet: @bluthg

- If you've already signed a CLA, it's possible we don't have your GitHub username or you're using a different email address. Please [contact us](https://icinga.com/company/contact/) if you think this is the case. - If you signed the CLA as a corporation, your GitHub username may not have been submitted to us. Please reach out to the responsible person in your organization.

[lippserd]

I'm a bit puzzled....

• What's the issue?
• Where's the test showing the issue?
• Do we really need this monstrosity of a function? I'm pretty sure there's a much nices (and less performance hungry!) way to get around this...

Hi Nick,

Hope you are well! Problem was/is something like this:

select tsrange('2021-10-28 11:43:08', '2021-10-28 13:02:52', '(]') - tsrange('2021-10-28 11:45:15', '2021-10-28 11:45:51', '[)');
ERROR:  result of range difference would not be contiguous

The first range is a period of time in which the object was down, and the second range is a downtime during this period.

I'm quite sure there's a better solution.

[cla-bot[bot]]

Thank you for your pull request. Before we can look at it, you'll need to sign a Contributor License Agreement (CLA).

Please follow instructions at https://icinga.com/company/contributor-agreement to sign the CLA.

After that, please reply here with a comment and we'll verify.

Contributors that have not signed yet: @bluthg

- If you've already signed a CLA, it's possible we don't have your GitHub username or you're using a different email address. Please [contact us](https://icinga.com/company/contact/) if you think this is the case. - If you signed the CLA as a corporation, your GitHub username may not have been submitted to us. Please reach out to the responsible person in your organization.

[bluthg]

Yeah, I figured that out :)

That's a tricky one indeed. I mean, the way the function is built right now does pass the tests I added. However they actually shouldn't pass...

postgres=# SELECT * from icinga_hoststatus where host_object_id = 7;
 host_object_id | status_update_time  | current_state | state_type
----------------+---------------------+---------------+------------
              7 | 2019-04-14 11:00:00 |             0 |          1
              7 | 2019-04-15 11:00:00 |             1 |          1
              7 | 2019-04-15 12:52:00 |             0 |          1
              7 | 2019-04-15 12:55:00 |             1 |          1
              7 | 2019-04-15 12:57:00 |             0 |          1
(5 rows)

postgres=# SELECT
postgres-#                 tsrange(actual_start_time, actual_end_time) AS downtime
postgres-#             FROM
postgres-#                 icinga_downtimehistory
postgres-#             WHERE
postgres-#                 object_id = 7
postgres-# ;
                   downtime
-----------------------------------------------
 ["2019-04-15 11:45:00","2019-04-15 11:46:00")
 ["2019-04-15 12:00:00","2019-04-15 12:05:00")
(2 rows)

postgres=# SELECT idoreports_get_sla_ok_percent(7,'2019-04-14 12:00:00','2019-04-15 15:00:00');
 idoreports_get_sla_ok_percent
-------------------------------
                           100
(1 row)

On top, just cutting a timerange in two will not cut it (pun intended) if you have TWO of these planned downtimes during one "server down" event. You'd end up with four overlapping time frames, I reckon.

[cla-bot[bot]]

Thank you for your pull request. Before we can look at it, you'll need to sign a Contributor License Agreement (CLA).

Please follow instructions at https://icinga.com/company/contributor-agreement to sign the CLA.

After that, please reply here with a comment and we'll verify.

Contributors that have not signed yet: @bluthg

- If you've already signed a CLA, it's possible we don't have your GitHub username or you're using a different email address. Please [contact us](https://icinga.com/company/contact/) if you think this is the case. - If you signed the CLA as a corporation, your GitHub username may not have been submitted to us. Please reach out to the responsible person in your organization.

[bluthg]

So, that last three commits (once the hellosign stops to fail, they should show up) should fix the issue without all the ado 😎

[cla-bot[bot]]

Thank you for your pull request. Before we can look at it, you'll need to sign a Contributor License Agreement (CLA).

Please follow instructions at https://icinga.com/company/contributor-agreement to sign the CLA.

After that, please reply here with a comment and we'll verify.

Contributors that have not signed yet: @bluthg

- If you've already signed a CLA, it's possible we don't have your GitHub username or you're using a different email address. Please [contact us](https://icinga.com/company/contact/) if you think this is the case. - If you signed the CLA as a corporation, your GitHub username may not have been submitted to us. Please reach out to the responsible person in your organization.

[bobapple]

@cla-bot check

[nilmerg]

Hey @bluthg,

many thanks for your changes! I don't know what this does or why it's better, but I see the increased efficiency which I really like :D

It works fine, too. :joy: