Make ClickhouseRepo task timeouts configurable in CE

[ruslandoga]

Changes

This PR makes CE's ClickhouseRepo tasks reuse timeouts defined on CLICKHOUSE_DATABASE_URL

For example, setting CLICKHOUSE_DATABASE_URL=http://plausible_events_db:8123/plausible_events_db?timeout=120000 would make the task timeout value equal eight minutes (480000=url_timeout*4). It has already been used to solve some timeout issues before: #3311 - so it makes sense to standardize on this approach rather than introduce extra ENV vars.

Related issues and discussions, where increasing the timeout value for tasks might help:

• Can't view stats for a longish time range #4257
• After updating from v2, API requests time out for longer time frames #4360
• Huge performance drop #4285

Some examples for how the timeout param in URL affects Repo config:

# for export CLICKHOUSE_DATABASE_URL=http://localhost:8123/plausible_events_db
iex> Keyword.fetch!(Plausible.ClickhouseRepo.config(), :timeout)
#==> 15000 <- this is the default from db_connection

# for export CLICKHOUSE_DATABASE_URL="http://localhost:8123/plausible_events_db?timeout=120000"
iex> Keyword.fetch!(Plausible.ClickhouseRepo.config(), :timeout)
#==> 120000 <- this would make tasks timeout after 120000*4=480000 or eight minutes

Tests

• Not sure how (or if) to test this, maybe config_test.exs can be updated, but for a custom repo like in Run migrations in order across repos #4466

Changelog

• Does not need a new entry, I'll just reply in the discussions above that the task timeouts can be configured in CLICKHOUSE_DATABASE_URL

Documentation

• This change does not need a documentation update

Dark mode

• This PR does not change the UI

[ruslandoga]

I'm doubling the timeout value from the repo just to make sure we don't timeout the task before db_connection does. It could be max(ch_timeout + 1, @task_timeout) but I thought that * 2 is safer since maybe some tasks do more than one DB request. Maybe it should be * 4 since this seems to be the default relation, (@task_timeout / default_db_connection_timeout = 60_000 / 15_000 = 4)

[zoldar]

Yeah, let's adjust it to 4 👍

[ruslandoga]

f331f58

[zoldar]

Not sure how (or if) to test this

@ruslandoga Perhaps starting a repo process without name and using https://hexdocs.pm/ecto/Ecto.Repo.html#c:put_dynamic_repo/1 would enable testing it (the original repo pid/name would have to be put back at the end of the test though).

[ruslandoga]

I tried adding the test:

• first attempt: f8675e0
• second attempt: f8fb13c

And now I think it is too complicated for what it is testing...
So I'm going to remove it once CI is green for the current iteration :) Done: ce960da