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

proposal: os: make the internal lockedfile package public #33974

Open
azr opened this issue Aug 30, 2019 · 30 comments
Open

proposal: os: make the internal lockedfile package public #33974

azr opened this issue Aug 30, 2019 · 30 comments
Labels
Milestone

Comments

@azr
Copy link
Contributor

azr commented Aug 30, 2019

An internal filelock pkg can be found here:
https://pkg.go.dev/cmd/go/internal/lockedfile/internal/filelock

A few projects are implementing file locking [1] but they do not seem to be maintained and I think they are not as nice as the internal filelock pkg I mentioned before. As a result some projects are doing their own version of it [2].

I suggest we make the filelock pkg public as I think this could be beneficial to the mass.

I'd be glad to do it given some pointers, like where to put it ?

Thanks !


[1] File lock projects:

https://github.com/gofrs/flock
https://github.com/MichaelS11/go-file-lock
https://github.com/juju/fslock

[2] Projects that implement their own file locking:

terraform

boltdb

@gopherbot gopherbot added this to the Proposal milestone Aug 30, 2019
@mvdan
Copy link
Member

mvdan commented Aug 30, 2019

Why not use the exported version at https://godoc.org/github.com/rogpeppe/go-internal? :)

@smasher164 smasher164 changed the title Proposal: make the internal filelock pkg public proposal: make the internal filelock package public Aug 30, 2019
@azr
Copy link
Contributor Author

azr commented Sep 3, 2019

Hey @mvdan, 🙂 here are a few reasons why I didn't go with it:

  • the filelock of go-internal is in an internal pkg
  • importing go-internal makes my go-deps bigger than necessary ( edit: this not such a big issue )
  • it feels wrong, if I find a bug, then I have to make a pull request to go and then wait for it to get merged and then wait for go-internals to update. It sounds like it has all the annoying effects of having these pkgs not internal anyway.

@mvdan
Copy link
Member

mvdan commented Sep 3, 2019

the filelock of go-internal is in an internal pkg

I'm not sure why that is, perhaps @rogpeppe or @myitcv can answer. Have you looked at https://godoc.org/github.com/rogpeppe/go-internal/lockedfile? That seems like a public, higher-level version of it.

That aside, sure, there are some upsides to having this in the standard library. But I think the downsides are generally greater:

  • Once a package is public, it can never have backwards incompatible changes
  • The standard library can grow, but never shrink in size
  • Once the number of users greatly outgrows just cmd/go, the maintainers will probably have extra work responding to issues, reviewing patches, and fixing bugs that might not even affect cmd/go

I think a far better proposal would be to include this in one of the official external repos, like x/sync. Once that's worked well for a while, then it can be part of the public standard library. This is the path that context took, for example.

@azr
Copy link
Contributor Author

azr commented Sep 3, 2019

lockedfile fits right to our usage and lockedfile doesn't have the filelock.ErrNotSupported. But yes lockedfile could work here.

Yes, I agree this should probably be in an x/ pkg 🙂, I only suggested we make the filelock pkg public and agree this would be a bit quick to put in the stdlib.

@ianlancetaylor
Copy link
Contributor

CC @bcmills

@bcmills
Copy link
Contributor

bcmills commented Sep 3, 2019

At this point we've been using lockedfile for a full release cycle and it seems mostly fine, modulo what appears to be an AIX kernel bug (#32817). I'd be fine with making it public.

Then the question is where to put it: x/sys, x/exp, x/sync, or someplace else entirely?

@azr
Copy link
Contributor Author

azr commented Sep 4, 2019

It's right at the intersection of sync and sys.
I think lockedfile should be in x/sync/lockedfile
I think that filelock is a bit lower level and could fit in x/sys/filelock ( if we ever make it public ). But these two are may be a bit redundant.

edit: regroup them both under x/exp/syssync ?

@bcmills
Copy link
Contributor

bcmills commented Sep 4, 2019

@azr, filelock should remain internal — it's much too subtle to use on its own. (In particular, if you want your program to remain portable, you have to be very careful to stick to a narrow subset of the possible modes of use.)

If folks discover other interesting use-cases for filelock, then we can probably add more ergonomic packages to address those use-cases while still keeping filelock internal.

@rsc
Copy link
Contributor

rsc commented Sep 25, 2019

Given that the author of filelock says it is not ready to be exported, it seems like this is a likely decline.

Leaving open for a week for final comments.

@bcmills
Copy link
Contributor

bcmills commented Sep 25, 2019

To be clear: I support making the lockedfile package itself public — I just want to ensure that the lockedfile/internal/filelock package remains internal to it.

@azr azr changed the title proposal: make the internal filelock package public proposal: make the internal lockedfile package public Sep 26, 2019
@azr
Copy link
Contributor Author

azr commented Sep 26, 2019

I just want to ensure that the lockedfile/internal/filelock package remains internal to it.

That's fine by me 🙂.

I think I'll put it in x/sync/lockedfile soonish; (unless you think another place is better ).

@rsc
Copy link
Contributor

rsc commented Oct 2, 2019

I misunderstood. Thanks for the clarification, @bcmills. Removing the FinalCommentPeriod label.

@rsc
Copy link
Contributor

rsc commented Oct 2, 2019

I don't think we are ready to add lockedfile to the standard library. It is a peculiar set of functions that does not really line up with the usual things in either os or io. Maybe it would be OK in x/sync but maybe it would be better to start in x/exp to understand if the API needs any adjustments.

If we are going to add a new public package (even in x/exp), I think the next step would be for someone to write a proper proposal doc laying out the API and explaining the decisions, get comments, and so on.

@azr, do you want to do that?

@azr
Copy link
Contributor Author

azr commented Oct 3, 2019

@rsc gotcha ! Yes I want to ! Where should such a proposal doc be written ? Is there a template for that ?

@bcmills
Copy link
Contributor

bcmills commented Oct 3, 2019

@azr, see https://github.com/golang/proposal#design-documents

@rsc
Copy link
Contributor

rsc commented Oct 9, 2019

Putting proposal on hold for design doc.

@azr
Copy link
Contributor Author

azr commented Oct 15, 2019

Hello there; I created the proposal ( first ever ! ) here golang/proposal#21; I hope it's good enough please feel free to give feedback 🙂 🙂 !

azr added a commit to azr/proposal that referenced this issue Oct 16, 2019
azr added a commit to azr/proposal that referenced this issue Oct 16, 2019
azr added a commit to azr/proposal that referenced this issue Oct 16, 2019
@ianlancetaylor
Copy link
Contributor

It's in the "incoming" list. We move proposals from "incoming" to "active" as we have bandwidth to handle them. See https://go.googlesource.com/proposal/+/refs/heads/master/README.md .

@arl
Copy link
Contributor

arl commented Jan 8, 2021

@ianlancetaylor I had missed that bit of information. Thank you

@tamilmani1989
Copy link

any update on making lockedfile package public?

@kensipe
Copy link

kensipe commented Feb 7, 2022

Some the details seem buried... The proposal process completed and landed the proposal on Mar 10, 2020: https://github.com/golang/proposal/blob/master/design/33974-add-public-lockedfile-pkg.md

The proposal looks good. It is unclear how / who to involve to move past proposal. Any insight @azr ? Happy to help if there is a need.

@ianlancetaylor
Copy link
Contributor

@kensipe This proposal has still not been accepted. I'll put it on the list for faster consideration.

@zpavlinovic
Copy link
Contributor

vuln/cmd/govulncheck could also make use of this library.

@rsc rsc changed the title proposal: make the internal lockedfile package public proposal: os: make the internal lockedfile package public Aug 10, 2022
@jastBytes
Copy link

Is it going to be public anytime soon?

dbussink added a commit to planetscale/vitess that referenced this issue Nov 16, 2022
The logic here is racy since multiple endtoend tests can create the file
at around the same time and allocate the same base port. This then leads
to flaky failing endtoend tests since not all ports can be used.

Even with the listening check later on, this is still racy because the
listen check might run at a later point still trying to allocate the
same port for the same service in different endtoend test.

This copies the Go filelock implementation, but only for unix systems
since that's the only supported platform for Vitess.

Go has a proposal open to make filelock available, but that is still
pending in golang/go#33974. Once that is
resolved, we can remove our copy of the implementation.

Signed-off-by: Dirkjan Bussink <d.bussink@gmail.com>
vmg pushed a commit to vitessio/vitess that referenced this issue Nov 16, 2022
* Fix race in determining start port for cluster

The logic here is racy since multiple endtoend tests can create the file
at around the same time and allocate the same base port. This then leads
to flaky failing endtoend tests since not all ports can be used.

Even with the listening check later on, this is still racy because the
listen check might run at a later point still trying to allocate the
same port for the same service in different endtoend test.

This copies the Go filelock implementation, but only for unix systems
since that's the only supported platform for Vitess.

Go has a proposal open to make filelock available, but that is still
pending in golang/go#33974. Once that is
resolved, we can remove our copy of the implementation.

Signed-off-by: Dirkjan Bussink <d.bussink@gmail.com>

* Change range of dynamic ports to not collide with tests

Next to the previous commit which fixes the race base port computation,
we also need to avoid collisions between ephemeral outgoing ports to
127.0.0.1 and the listening ones.

The problem is that we set 1024 as the first ephemeral port, but that
means we can collide and trigger flaky tests with all the tests that
allocate a server port as well.

Given we start allocating server ports now starting at 6700 for endtoend
tests and we allocate in blocks of 200, we have room for 75+ concurrent
endtoend tests which should totally suffice.

We still have a range of 40+ ephemeral ports which should still be
sufficient as well.

Signed-off-by: Dirkjan Bussink <d.bussink@gmail.com>

Signed-off-by: Dirkjan Bussink <d.bussink@gmail.com>
jevolk added a commit to jevolk/nats-server that referenced this issue Aug 18, 2023
Running two instances of the server which share the same directory (e.g.
default configuration `/tmp/nats/jetstream') will corrupt each other.

We mitigate by creating an empty file called LOCK in the directory and then
acquire a posix `flock(2)` on it; a common pattern.

Note this is not inherently cross-platform without golang/go#33974

Signed-off-by: Jason Volk <jason@zemos.net>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
Projects
Status: Incoming
Development

No branches or pull requests