Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Change with_tokio_rt to accept Arc<Runtime>. #599

Draft
wants to merge 1 commit into
base: master
Choose a base branch
from

Conversation

gz
Copy link

@gz gz commented Sep 4, 2024

This allows to share tokio runtimes across different sub-systems inside your application.

PR Type

Feature, Refactor

#580

PR Checklist

Check your PR fulfills the following:

  • Tests for the changes have been added / updated.
  • Documentation comments have been added / updated.
  • A changelog entry has been made for the appropriate packages.
  • Format code with the latest stable rustfmt

Overview

The current signature of https://docs.rs/actix-rt/latest/actix_rt/struct.System.html#method.with_tokio_rt expects a Runtime. It would be better if I can supply an Arc instead. The reason being that is that in a bigger application we have multiple systems trying to leverage tokio. With the given API I still need to create a separate runtime for actix. Having the same runtime is helpful for better control over CPU resources as there aren't n runtimes competing for CPU time.

FWIW this is a breaking change so if there alternative ways to achieve this I'm happy to give those a try too.

This allows to share tokio runtimes across different sub-systems
inside your application.

Signed-off-by: Gerd Zellweger <[email protected]>
@gz gz marked this pull request as draft September 4, 2024 22:18
gz added a commit to feldera/feldera that referenced this pull request Sep 5, 2024
If an adhoc query is sent it will be handled in the same
tokio runtime that is running the webserver.

This raises the questions of how many cores the runtime should
have. Datafusion will need one task for each partition we're
reading from. And we have as many partitions as dbsp workers.
So maybe we should've just change HTTP_NUM_WORKERS from 4 to
`config.global.runtime_workers`. Except, this doesn't
quite help because if your query is very expensive like
`select * from table_with_100m`, all it's task run for
quite some time and starve out the HTTP request. So now
the UI no longer updates. Ok, so maybe it should
just be `workers*2`, but if workers is set near the amount
of CPU cores on the machine then we'll overprovision
the tokio runtime and that will slow things down too.
So for now we just set this to #cores on the machine by
not configuring it (this is the default).

Note aside, ideally we can just swap out the actix
runtime with `dbsp::runtime::TOKIO` which is what we use
everywhere else and this already uses the #cores
in the runtime to run the tasks.
However, this also doesn't work because the actix
API has a stupid bug in the interface.

I submitted a PR to fix this here:
actix/actix-net#599

So for now, we'll just have two tokio runtimes, with

Signed-off-by: Gerd Zellweger <[email protected]>
gz added a commit to feldera/feldera that referenced this pull request Sep 7, 2024
If an adhoc query is sent it will be handled in the same
tokio runtime that is running the webserver.

This raises the questions of how many cores the runtime should
have. Datafusion will need one task for each partition we're
reading from. And we have as many partitions as dbsp workers.
So maybe we should've just change HTTP_NUM_WORKERS from 4 to
`config.global.runtime_workers`. Except, this doesn't
quite help because if your query is very expensive like
`select * from table_with_100m`, all it's task run for
quite some time and starve out the HTTP request. So now
the UI no longer updates. Ok, so maybe it should
just be `workers*2`, but if workers is set near the amount
of CPU cores on the machine then we'll overprovision
the tokio runtime and that will slow things down too.
So for now we just set this to #cores on the machine by
not configuring it (this is the default).

Note aside, ideally we can just swap out the actix
runtime with `dbsp::runtime::TOKIO` which is what we use
everywhere else and this already uses the #cores
in the runtime to run the tasks.
However, this also doesn't work because the actix
API has a stupid bug in the interface.

I submitted a PR to fix this here:
actix/actix-net#599

So for now, we'll just have two tokio runtimes, with

Signed-off-by: Gerd Zellweger <[email protected]>
gz added a commit to feldera/feldera that referenced this pull request Sep 7, 2024
If an adhoc query is sent it will be handled in the same
tokio runtime that is running the webserver.

This raises the questions of how many cores the runtime should
have. Datafusion will need one task for each partition we're
reading from. And we have as many partitions as dbsp workers.
So maybe we should've just change HTTP_NUM_WORKERS from 4 to
`config.global.runtime_workers`. Except, this doesn't
quite help because if your query is very expensive like
`select * from table_with_100m`, all it's task run for
quite some time and starve out the HTTP request. So now
the UI no longer updates. Ok, so maybe it should
just be `workers*2`, but if workers is set near the amount
of CPU cores on the machine then we'll overprovision
the tokio runtime and that will slow things down too.
So for now we just set this to #cores on the machine by
not configuring it (this is the default).

Note aside, ideally we can just swap out the actix
runtime with `dbsp::runtime::TOKIO` which is what we use
everywhere else and this already uses the #cores
in the runtime to run the tasks.
However, this also doesn't work because the actix
API has a stupid bug in the interface.

I submitted a PR to fix this here:
actix/actix-net#599

So for now, we'll just have two tokio runtimes, with

Signed-off-by: Gerd Zellweger <[email protected]>
gz added a commit to feldera/feldera that referenced this pull request Sep 7, 2024
If an adhoc query is sent it will be handled in the same
tokio runtime that is running the webserver.

This raises the questions of how many cores the runtime should
have. Datafusion will need one task for each partition we're
reading from. And we have as many partitions as dbsp workers.
So maybe we should've just change HTTP_NUM_WORKERS from 4 to
`config.global.runtime_workers`. Except, this doesn't
quite help because if your query is very expensive like
`select * from table_with_100m`, all it's task run for
quite some time and starve out the HTTP request. So now
the UI no longer updates. Ok, so maybe it should
just be `workers*2`, but if workers is set near the amount
of CPU cores on the machine then we'll overprovision
the tokio runtime and that will slow things down too.
So for now we just set this to #cores on the machine by
not configuring it (this is the default).

Note aside, ideally we can just swap out the actix
runtime with `dbsp::runtime::TOKIO` which is what we use
everywhere else and this already uses the #cores
in the runtime to run the tasks.
However, this also doesn't work because the actix
API has a stupid bug in the interface.

I submitted a PR to fix this here:
actix/actix-net#599

So for now, we'll just have two tokio runtimes, with

Signed-off-by: Gerd Zellweger <[email protected]>
@robjtede
Copy link
Member

To avoid breaks we could add a new method for this.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

2 participants