mxnet-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Marco de Abreu <marco.g.ab...@gmail.com>
Subject Re: CI Pipeline Change Proposal
Date Fri, 27 Mar 2020 14:57:10 GMT
The docker cache images can be used by you. They're available in Dockerhub,
you just have to tweak the docker run method.

The thing is that the scripts CI uses has the intention that layers change
and thus the cache is used.

If you want to be able to change the layers, then you have to accept the
fact that docker build is required. If you are fine with accepting whatever
is published in Dockerhub, then you can use docker run.

The thing which is missing here is a flag in build.py which states whether
either the local dir or the remote cache should be used to create the
docker environment. Your dir would still be mounted into the container, but
the difference is that the layers would not necessarily match to what's in
your dir. Speak, if you change the layers scripts, they'd obviously not be
available in the image if you consume Dockerhub.

Would such a feature help you reduce the Painpoints you are encountering?

With regards to the pinning: I think it is not feasible to not do pinning
but at the same time expect everything to stay the same. I can recommend
tackling that concern by introducing pinning and having an automated system
(or an assigned person) which tests later versions on a repeating base.

-Marco

Aaron Markham <aaron.s.markham@gmail.com> schrieb am Fr., 27. März 2020,
15:48:

> Sure. That's the fix for now.
>
> But, I've noticed that when that's done and there's no process to enforce
> upgrades and patching, these get really out of date and the problems
> compound.
>
> Plus, when I build locally using docker, I can never seem to get the
> benefit of the cache. Or at least not in the way I'd expect it to be.
>
> On one hand I get to discover all these bugs before they hit prod. Ha.
> But I'd like to have dockerfiles that pull base images that have v1.6.0 and
> each patched major release. Ideally I'd have ones for each language binding
> too.
> Seems like that might really simplify trying to work with a particular
> issue. Instead, I'm constantly rebuilding binaries and having to reset the
> submodules and make clean.
> Seems even more relevant now that we're maintaining master and 1.7.x and at
> least for me, 1.6.x.
>
>
>
> On Fri, Mar 27, 2020, 00:55 Marco de Abreu <marco.g.abreu@gmail.com>
> wrote:
>
> > What about dependency pinning?
> >
> > The cache should not be our method to do dependency pinning and
> > synchronization.
> >
> > -Marco
> >
> > Aaron Markham <aaron.s.markham@gmail.com> schrieb am Fr., 27. März 2020,
> > 03:45:
> >
> > > I'm dealing with a Ruby dep breaking the site build right now.
> > > I wish this would be on occasion that I choose, not when Ruby or x
> > > dependency releases a new version. When the cache expires for Jekyll
> the
> > > site won't publish anymore... And CI will be blocked for the website
> > test.
> > >
> > > If we built the base OS and main deps once when we do a minor release,
> > > upload that to dockerhub, then we'd save build time and things breaking
> > > randomly. Users can use those docker images too. At release time we do
> a
> > > round of updates and testing when we're ready. Can we find a balance
> > > between caching, prebuilt docker images, freshness, and efficiency?
> > >
> > >
> > > On Thu, Mar 26, 2020, 14:31 Marco de Abreu <marco.g.abreu@gmail.com>
> > > wrote:
> > >
> > > > Correct. But I'm surprised about 2:50min to pull down the images.
> > > >
> > > > Maybe it makes sense to use ECR as mirror?
> > > >
> > > > -Marco
> > > >
> > > > Joe Evans <joseph.evans@gmail.com> schrieb am Do., 26. März 2020,
> > 22:02:
> > > >
> > > > > +1 on rebuilding the containers regularly without caching layers.
> > > > >
> > > > > We are both pulling down a bunch of docker layers (when docker
> pulls
> > an
> > > > > image) and then building a new container to run the sanity build
> in.
> > > > > Pulling down all the layers is what is taking so long (2m50s.)
> Within
> > > the
> > > > > docker build, all the layers are cached, so it doesn't take long.
> > > Unless
> > > > > I'm missing something, it doesn't make much sense to be rebuilding
> > the
> > > > > image every build.
> > > > >
> > > > > On Thu, Mar 26, 2020 at 1:12 PM Lausen, Leonard
> > > > <lausen@amazon.com.invalid
> > > > > >
> > > > > wrote:
> > > > >
> > > > > > WRT Docker Cache: We need to add a mechanism to invalidate the
> > cache
> > > > and
> > > > > > rebuild
> > > > > > the containers on a set schedule. The builds break too often
and
> > the
> > > > > > breakage is
> > > > > > only detected when a contributor touches the Dockerfiles
> (manually
> > > > > causing
> > > > > > cache
> > > > > > invalidation)
> > > > > >
> > > > > > On Thu, 2020-03-26 at 16:06 -0400, Aaron Markham wrote:
> > > > > > > I think it is a good idea to do the sanity check first.
Even at
> > 10
> > > > > > minutes.
> > > > > > > And also try to fix the docker cache situation, but those
can
> be
> > > > > separate
> > > > > > > tasks.
> > > > > > >
> > > > > > > On Thu, Mar 26, 2020, 12:52 Marco de Abreu <
> > > marco.g.abreu@gmail.com>
> > > > > > wrote:
> > > > > > >
> > > > > > > > Jenkins doesn't load for me, so let me ask this way:
are we
> > > > actually
> > > > > > > > rebuilding every single time or do you mean the docker
cache?
> > > > Pulling
> > > > > > the
> > > > > > > > cache should only take a few seconds from my experience
-
> > docker
> > > > > build
> > > > > > > > should be a no-op in most cases.
> > > > > > > >
> > > > > > > > -Marco
> > > > > > > >
> > > > > > > >
> > > > > > > > Joe Evans <joseph.evans@gmail.com> schrieb am
Do., 26. März
> > > 2020,
> > > > > > 20:46:
> > > > > > > >
> > > > > > > > > The sanity-lint check pulls a docker image cache,
builds a
> > new
> > > > > > container
> > > > > > > > > and runs inside. The docker setup is taking around
3
> minutes,
> > > at
> > > > > > least:
> > > > > > > > >
> > > > > > > > >
> > > > > > > > >
> > > > > > > >
> > > > > >
> > > > >
> > > >
> > >
> >
> http://jenkins.mxnet-ci.amazon-ml.com/blue/organizations/jenkins/mxnet-validation%2Fsanity/detail/master/1764/pipeline/39
> > > > > > > > > We could improve this by not having to build
a new
> container
> > > > every
> > > > > > time.
> > > > > > > > > Also, our CI containers are huge so it takes
awhile to pull
> > > them
> > > > > > down.
> > > > > > > > I'm
> > > > > > > > > sure we could reduce the size by being a bit
more careful
> in
> > > > > building
> > > > > > > > them
> > > > > > > > > too.
> > > > > > > > >
> > > > > > > > > Joe
> > > > > > > > >
> > > > > > > > > On Thu, Mar 26, 2020 at 12:33 PM Marco de Abreu
<
> > > > > > marco.g.abreu@gmail.com
> > > > > > > > >
> > > > > > > > > wrote:
> > > > > > > > >
> > > > > > > > > > Do you know what's driving the duration
for sanity? It
> used
> > > to
> > > > be
> > > > > > 50
> > > > > > > > sec
> > > > > > > > > > execution and 60 sec preparation.
> > > > > > > > > >
> > > > > > > > > > -Marco
> > > > > > > > > >
> > > > > > > > > > Joe Evans <joseph.evans@gmail.com>
schrieb am Do., 26.
> > März
> > > > > 2020,
> > > > > > > > 20:31:
> > > > > > > > > > > Thanks Marco and Aaron for your input.
> > > > > > > > > > >
> > > > > > > > > > > > Can you show by how much the duration
will increase?
> > > > > > > > > > >
> > > > > > > > > > > The average sanity build time is around
10min, while
> the
> > > > > average
> > > > > > > > build
> > > > > > > > > > time
> > > > > > > > > > > for unix-cpu is about 2 hours, so the
entire build
> > pipeline
> > > > > would
> > > > > > > > > > increase
> > > > > > > > > > > by 2 hours if we required both unix-cpu
and sanity to
> > > > complete
> > > > > in
> > > > > > > > > > parallel.
> > > > > > > > > > > I took a look at the CloudWatch metrics
we're saving
> for
> > > > > Jenkins
> > > > > > > > jobs.
> > > > > > > > > > Here
> > > > > > > > > > > is the failure rate per job, based
on builds triggered
> by
> > > PRs
> > > > > in
> > > > > > the
> > > > > > > > > past
> > > > > > > > > > > year. As you can see, the sanity build
failure is still
> > > > fairly
> > > > > > high
> > > > > > > > and
> > > > > > > > > > > would save a lot of unneeded build
jobs.
> > > > > > > > > > >
> > > > > > > > > > > Job Successful Failed Failure Rate
> > > > > > > > > > > sanity 6900 2729 28.34%
> > > > > > > > > > > unix-cpu 4268 4786 52.86%
> > > > > > > > > > > unix-gpu 3686 5637 60.46%
> > > > > > > > > > > centos-cpu 6777 2809 29.30%
> > > > > > > > > > > centos-gpu 6318 3350 34.65%
> > > > > > > > > > > clang 7879 1588 16.77%
> > > > > > > > > > > edge 7654 1933 20.16%
> > > > > > > > > > > miscellaneous 8090 1510 15.73%
> > > > > > > > > > > website 7226 2179 23.17%
> > > > > > > > > > > windows-cpu 6084 3621 37.31%
> > > > > > > > > > > windows-gpu 5191 4721 47.63%
> > > > > > > > > > >
> > > > > > > > > > > We can start by requiring only the
sanity job to
> complete
> > > > > before
> > > > > > > > > > triggering
> > > > > > > > > > > the rest, and collect data to decide
if it makes sense
> to
> > > > > change
> > > > > > it
> > > > > > > > > from
> > > > > > > > > > > there. Any objections to this approach?
> > > > > > > > > > >
> > > > > > > > > > > Thanks.
> > > > > > > > > > > Joe
> > > > > > > > > > >
> > > > > > > > > > >
> > > > > > > > > > > On Wed, Mar 25, 2020 at 9:35 AM Marco
de Abreu <
> > > > > > > > > marco.g.abreu@gmail.com>
> > > > > > > > > > > wrote:
> > > > > > > > > > >
> > > > > > > > > > > > Back then I have created a system
which exports all
> > > Jenkins
> > > > > > results
> > > > > > > > > to
> > > > > > > > > > > > cloud watch. It does not include
individual test
> > results
> > > > but
> > > > > > rather
> > > > > > > > > > > stages
> > > > > > > > > > > > and jobs. The data for the sanity
check should be
> > > available
> > > > > > there.
> > > > > > > > > > > >
> > > > > > > > > > > > Something I'd also be curious
about is the percentage
> > of
> > > > the
> > > > > > > > failures
> > > > > > > > > > in
> > > > > > > > > > > > one run. Speak, if a commit failed,
have there been
> > > > multiple
> > > > > > jobs
> > > > > > > > > > failing
> > > > > > > > > > > > (indicating an error in the code)
or only one or two
> > > > > > (indicating
> > > > > > > > > > > > flakyness). This should give us
a proper
> understanding
> > of
> > > > how
> > > > > > > > > > unnecessary
> > > > > > > > > > > > these runs really are.
> > > > > > > > > > > >
> > > > > > > > > > > > -Marck
> > > > > > > > > > > >
> > > > > > > > > > > > Aaron Markham <aaron.s.markham@gmail.com>
schrieb am
> > > Mi.,
> > > > > 25.
> > > > > > März
> > > > > > > > > > 2020,
> > > > > > > > > > > > 16:53:
> > > > > > > > > > > >
> > > > > > > > > > > > > +1 for sanity check - that's
fast.
> > > > > > > > > > > > > -1 for unix-cpu - that's
slow and can just hang.
> > > > > > > > > > > > >
> > > > > > > > > > > > > So my suggestion would be
to see the data apart -
> > > what's
> > > > > the
> > > > > > > > > failure
> > > > > > > > > > > > > rate on the sanity check
and the unix-cpu?
> Actually,
> > > can
> > > > we
> > > > > > get a
> > > > > > > > > > > > > table of all of the tests
with this data?!
> > > > > > > > > > > > > If the sanity check fails...
let's say 20% of the
> > time,
> > > > but
> > > > > > only
> > > > > > > > > > takes
> > > > > > > > > > > > > a couple of minutes, then
ya, let's stack it and do
> > > that
> > > > > one
> > > > > > > > first.
> > > > > > > > > > > > > I think unix-cpu needs to
be broken apart. It's too
> > > > complex
> > > > > > and
> > > > > > > > > fails
> > > > > > > > > > > > > in multiple ways. Isolate
the brittle parts. Then
> we
> > > can
> > > > > > > > > > > > > restart/disable those as
needed, while all of the
> > other
> > > > > parts
> > > > > > > > pass
> > > > > > > > > > and
> > > > > > > > > > > > > don't have to be rerun.
> > > > > > > > > > > > >
> > > > > > > > > > > > > On Wed, Mar 25, 2020 at 1:32
AM Marco de Abreu <
> > > > > > > > > > > marco.g.abreu@gmail.com>
> > > > > > > > > > > > > wrote:
> > > > > > > > > > > > > > We had this structure
in the past and the
> community
> > > was
> > > > > > > > bothered
> > > > > > > > > by
> > > > > > > > > > > CI
> > > > > > > > > > > > > > taking more time, thus
we moved to the current
> > model
> > > > with
> > > > > > > > > > everything
> > > > > > > > > > > > > > parallelized. We'd basically
revert that then.
> > > > > > > > > > > > > >
> > > > > > > > > > > > > > Can you show by how
much the duration will
> > increase?
> > > > > > > > > > > > > >
> > > > > > > > > > > > > > Also, we have zero test
parallelisation, speak we
> > are
> > > > > > running
> > > > > > > > one
> > > > > > > > > > > test
> > > > > > > > > > > > on
> > > > > > > > > > > > > > 72 core machines (although
multiple workers).
> > > Wouldn't
> > > > it
> > > > > > be
> > > > > > > > way
> > > > > > > > > > more
> > > > > > > > > > > > > > efficient to add parallelisation
and thus heavily
> > > > reduce
> > > > > > the
> > > > > > > > time
> > > > > > > > > > > spent
> > > > > > > > > > > > > on
> > > > > > > > > > > > > > the tasks instead of
staggering?
> > > > > > > > > > > > > >
> > > > > > > > > > > > > > I feel concerned that
these measures to save cost
> > are
> > > > > paid
> > > > > > in
> > > > > > > > the
> > > > > > > > > > > form
> > > > > > > > > > > > > of a
> > > > > > > > > > > > > > worse user experience.
I see a big potential to
> > save
> > > > > costs
> > > > > > by
> > > > > > > > > > > > increasing
> > > > > > > > > > > > > > efficiency while actually
improving the user
> > > experience
> > > > > > due to
> > > > > > > > CI
> > > > > > > > > > > being
> > > > > > > > > > > > > > faster.
> > > > > > > > > > > > > >
> > > > > > > > > > > > > > -Marco
> > > > > > > > > > > > > >
> > > > > > > > > > > > > > Joe Evans <joseph.evans@gmail.com>
schrieb am
> Mi.,
> > > 25.
> > > > > > März
> > > > > > > > > 2020,
> > > > > > > > > > > > 04:58:
> > > > > > > > > > > > > > > Hi,
> > > > > > > > > > > > > > >
> > > > > > > > > > > > > > >
> > > > > > > > > > > > > > > First, I just wanted
to introduce myself to the
> > > MXNet
> > > > > > > > > community.
> > > > > > > > > > > I’m
> > > > > > > > > > > > > Joe
> > > > > > > > > > > > > > > and will be working
with Chai and the AWS team
> to
> > > > > improve
> > > > > > > > some
> > > > > > > > > > > issues
> > > > > > > > > > > > > > > around MXNet CI.
One of our goals is to reduce
> > the
> > > > > costs
> > > > > > > > > > associated
> > > > > > > > > > > > > with
> > > > > > > > > > > > > > > running MXNet CI.
The task I’m working on now
> is
> > > this
> > > > > > issue:
> > > > > > > > > > > > > > >
> > > > > > > > > > > > > > >
> > > > > > > > > > > > > > >
> > > > https://github.com/apache/incubator-mxnet/issues/17802
> > > > > > > > > > > > > > >
> > > > > > > > > > > > > > >
> > > > > > > > > > > > > > > Proposal: Staggered
Jenkins CI pipeline
> > > > > > > > > > > > > > >
> > > > > > > > > > > > > > >
> > > > > > > > > > > > > > > Based on data collected
from Jenkins, around
> 55%
> > of
> > > > the
> > > > > > time
> > > > > > > > > when
> > > > > > > > > > > the
> > > > > > > > > > > > > > > mxnet-validation
CI build is triggered by a PR,
> > > > either
> > > > > > the
> > > > > > > > > sanity
> > > > > > > > > > > or
> > > > > > > > > > > > > > > unix-cpu builds
fail. When either of these
> builds
> > > > fail,
> > > > > > it
> > > > > > > > > > doesn’t
> > > > > > > > > > > > make
> > > > > > > > > > > > > > > sense to run the
rest of the pipelines and
> > utilize
> > > > all
> > > > > > those
> > > > > > > > > > > > resources
> > > > > > > > > > > > > if
> > > > > > > > > > > > > > > we’ve already
identified a build or unit test
> > > > failure.
> > > > > > > > > > > > > > >
> > > > > > > > > > > > > > >
> > > > > > > > > > > > > > > We are proposing
changing the MXNet Jenkins CI
> > > > pipeline
> > > > > > by
> > > > > > > > > > > requiring
> > > > > > > > > > > > > the
> > > > > > > > > > > > > > > *sanity* and *unix-cpu*
builds to complete and
> > pass
> > > > > tests
> > > > > > > > > > > > successfully
> > > > > > > > > > > > > > > before starting
the other build pipelines
> > > > > > (centos-cpu/gpu,
> > > > > > > > > > > unix-gpu,
> > > > > > > > > > > > > > > windows-cpu/gpu,
etc.) Once the sanity builds
> > > > > > successfully
> > > > > > > > > > > complete,
> > > > > > > > > > > > > the
> > > > > > > > > > > > > > > remaining build
pipelines will be triggered and
> > run
> > > > in
> > > > > > > > parallel
> > > > > > > > > > (as
> > > > > > > > > > > > > they
> > > > > > > > > > > > > > > currently do.)
The purpose of this change is to
> > > > > identify
> > > > > > > > faulty
> > > > > > > > > > > code
> > > > > > > > > > > > or
> > > > > > > > > > > > > > > compatibility issues
early and prevent further
> > > > > execution
> > > > > > of
> > > > > > > > CI
> > > > > > > > > > > > builds.
> > > > > > > > > > > > > This
> > > > > > > > > > > > > > > will increase the
time required to test a PR,
> but
> > > > will
> > > > > > > > prevent
> > > > > > > > > > > > > unnecessary
> > > > > > > > > > > > > > > builds from running.
> > > > > > > > > > > > > > >
> > > > > > > > > > > > > > >
> > > > > > > > > > > > > > > Does anyone have
any concerns with this change
> or
> > > > > > > > suggestions?
> > > > > > > > > > > > > > >
> > > > > > > > > > > > > > > Thanks.
> > > > > > > > > > > > > > >
> > > > > > > > > > > > > > > Joe Evans
> > > > > > > > > > > > > > >
> > > > > > > > > > > > > > > joseph.evans@gmail.com
> > > > > > > > > > > > > > >
> > > > > >
> > > > >
> > > >
> > >
> >
>

Mime
  • Unnamed multipart/alternative (inline, None, 0 bytes)
View raw message