►
From YouTube: GitLab 12.5 Kickoff - Runner
Description
A 5 minute overview of what we plan to deliver for the Runner team in our 12.5 milestone, shipping on Nov 22nd 2019.
A
Hi
I'm
Josh
Lambert
product
manager
here
at
get
lab,
and
today
I'd
like
to
give
you
more
information
on
the
issues
and
features
and
bugs
that
were
working
on
in
our
1205
milestone
for
our
get
lab
on
our
team.
This
release
is
planned
to
be
available
on
November
22nd
and
we
can't
wait
to
deliver
these
features
to
you.
The
first
thing
we're
working
on
is
a
number
of
the
high
priority
bugs
that
we
have
and
then
the
second
item
that
we're
working
on
is
that
we'd
like
to
deliver
to
you.
A
The
first
iteration
of
our
windows
shared
our
windows.
Auto
scaling
is
I.
Could
our
four
VMs?
This
would
be
the
first
auto
scaling
executor
that
github
has
that
can
support
Windows
and
so
we're
very
excited
about
that,
and
it
will
also
support
us
to
offer
shared
runners
on
our
get
live.com
service
as
well.
So
let's
look
through
the
issues
here
in
more
detail.
The
first
one
is
as
a
carryover
from
our
prior
release,
where
it's
a
bug
that,
in
some
cases,
jobs
that.
A
Mning
terminated
before
they're
completed
and
when
that
happens,
actually
it
reported
as
success
back
to
the
platform
as
I
post
their
actual
end
state,
and
this
can
lead
to
jobs
that
actually
should
have
failed,
but
affect
your
reported
as
success
back
up
to
the
platform,
and
that
can
cause
the
pipeline
to
continue
when
it
shouldn't.
We
try
to
solve
this
again
in
the
last
milestone.
We
thought
we
had
to
fix.
We
delivered
it,
it
really
released
to
our
community
and
they
gratefully
did
some
testing
and
confirmed
that.
Actually,
the
problem
still
existed.
A
So
we're
essentially
back
to
digging
into
further
details
on
this
issue,
but
we
do
want
to
solve
it
and
I
know
how
important
it
is
for
you.
The
next
item
is
another
limitation
that
we
found
actually
in
this
time
in
docker
machine.
So
our
doctor
machines,
Eckener
in
some
cases,
can
consume
a
number
of
EPI
queries
depending
on
your
scale
in
the
API
limits
new
project.
A
When
this
happens
and
you
hit
the
cap,
essentially
you
start
to
become
throttled
and
what
happens
here
is
that
you
can
have
machines
that
fail
to
start
or
machines
that
are
failed
to
clean
up
once
the
job
is
over
the
leading
to
access
compute
usage
utilized
in
your
platform.
So
with
this
fix
here
on
AWS
we're
actually
going
to
be
shipping.
A
forked
version
attacker
machine.
This
we
catch
doctor
machine
is
currently
in
maintenance
mode
and
there
isn't
a
clear
path
to
win
or
if
things
would
get
merged
upstream
from
the
community.
A
We're
also
working
to
fix,
buying
publication,
no
longer
being
valid
for
some
vine
wedding
and
also
we're
going
to
working
to
address
another
carryover
from
12.4,
which
we
didn't
get
a
chance
to
address
last
release,
which
is
that
in
some
cases,
long-running
jobs
are
canceled
in
AU
I,
don't
actually
interrupt.
The
job
is
in
progress,
and
so
in
this
case,
for
example,
if
you
use-
and
you
interrupt
a
little
keyword
in
or
if
you
just
the
middle
cancel
button,
the
job
could
continue
to
run,
which
is
misleading.
A
Him
obviously
defeats
the
purpose
of
the
interruptible
flag.
So
we
want
to
go
ahead
and
get
that
fixed
for
you
as
well.
We're
also
working
to
fix
an
issue
with
the
that
was
introduced
with
the
file
locking
and
the
configuration
file
for,
for
our
runners
in
particular,
on
Windows
Windows
is
a
bit
of
different
file.
A
Locking
strategy
in
our
Linux
runner,
stuff
and
one
of
Linux
platform
does
and
it's
entirely
exclusive,
and
so
this
was
leading
to
some
problems
and
we're
looking
to
address
those
lifee
in
this
first
iteration
here
by
just
not
doing
a
file
walking
and
we're
bringing
back
the
old
behavior
on
the
Windows
platform.
And
then
we
get
us
more
time
to
go
through
and
try
to
have
a
an
improved
locking
of
the
configuration
file
without
being
totally
exclusive.
A
The
next
set
of
issues
are
all
the
remaining
work
for
us
to
be
able
to
deliver
that
Windows
auto-scaling
executor
for
both
GCP
and
hopefully,
as
a
stretch,
goal
Azure
as
well.
So
again,
we're
very
excited
about
this.
It'll
be
the
first
auto-scaling
runner
that
we
have
available
that
could
support,
Windows,
builds
and
so
we're
looking
forward
to
delivering
this
to
you
and
getting
into
your
hands.
So
that's
what
we
have
planned
for
our
12.5
release.
A
We're
very
excited
to
again
address
some
of
these
bugs,
as
well
as
the
first
version
of
our
auto-scaling
runner
for
Windows,
and
we
can't
wait
to
get
it
into
your
hands
and
hear
what
you
have
to
say
about
it.
So
please
participate
on
our
issues.
Open
up,
new
ones
for
bugs
might
find
or
features
are
looking
for
and,
of
course,
always
vote
music
Anna
comment
on
our
current
plants.
Thank
you
so
much
and
looking
forward
to
talking
to
you
further
and
our
next
release.