►
From YouTube: 20191202 Cluster API Provider AWS Office Hours
Description
20191202 Cluster API Provider AWS Office Hours
A
Please
do
be
aware
that
you
know
that
this
meeting
is
conforming
with
the
kubernetes
community
guidelines,
so
in
general,
please
be
excellent
to
one
another.
If
you
have
any
items
that
you
would
like
to
add,
please
go
ahead
and
add
them
to
the
agenda
and
I
will
go
ahead
and
link
that
in
the
chat
as
well.
This
meeting
is
also
recorded
alright
and
also,
if
you
can
add
yourself
to
the
attending
list,
that
would
be
great
as
well.
A
To
start
with,
we
have
a
couple
of
PSAs
that
there
have
been
a
couple
of
releases
of
the
uws
provider
since
our
last
meeting,
0.4,
top-five
and
0.4.
That
states
have
both
been
released
generally.
They
are
bug
fixes,
with
the
exception
of
one
new
feature,
which
is
support
for
cluster
names
that
are
greater
than
22
characters.
B
B
A
A
B
C
B
C
Yeah
I
think
we
need
to
do
the
metric
stuff
first
yeah.
This
is
good.
It's
bit
hard
to
base
need
information
about
when
it's
happening.
First
I
mean
it's
okay.
They
would
like
to
do
some
more
doing.
I
think
it
does
eventually
do
the
right
thing
anyway,
at
the
moment,
so
I
don't
think
it's
MC
per
high
priority
right
now.
Okay,.
B
That's
fair,
so
I
will
I'll
be
optimistic
and
stick
it
in
the
next
release,
but
I
will
go
with
long-term,
given
your
statement
that
it
most
likely
eventually
resolves
itself
sound,
fair,
yes,
okay,
cool
thanks
all
right,
Chuck
open
one
called
deletes
are
too
slow.
I,
don't
really
think,
there's
anything
we
can
do
here.
It's
been
like
this
for
a
while
sometime
instances
take
multiple
minutes
to
get
to
terminated.
Releasing
the
elastic
IP
ease
to
leave
in
that
gateway
is
Internet
gateways
just
take
a
while
is.
B
A
It's
kind
of
tricky
because
we
do
have
a
default
parallelization
now
of
greater
than
one,
so
we
are
doing
more
of
these
operations
in
parallel
than
we
previously
were
doing,
but
once
we
hit
kind
of
the
max
parallelization,
we
do
kind
of
block
on
some
of
these
operations.
Instead
of
kind
of
handling
them
were
asynchronously,
so
we
can
potentially
add
some
asynchronous
handling
around
the
termination
of
instances
and
some
of
the
deletions
that
some
of
the
deletion
components
that
take
a
bit
of
time
right
now
and
that
would
help
with
kind
of
longer-term
scalability.
A
B
B
B
B
B
B
D
D
It,
for
example,
I
had
a
case
where
I
had
the
wrong
permissions
or
Vista
was
missing
one
permission
where
he'll
be
living
couldn't
get
deployed
and
the
deletion
it
will
also
get
stuck
at
deleting
the
it
will
be
because
again
didn't
have
information,
so
that
gives
a
bit
of
a
robustness
issue
in
the
deletion.
So
my
proposal
was
basically
to
at
some
point
my
process
to
remove
the
synchronous,
asynchronous
way
of
deleting
are
just
trying
to
delete
everything
and
then,
at
the
end,
gathered
the
errors
and
see
if
I
skip
all
things
got
immediately.
D
C
D
C
D
B
D
D
C
B
So
I
think
it's
probably
a
good
idea
to
try
to
make
some
improvements
here
in
the
next
release,
mainly
because
you're,
not
the
only
one
or
one
who's
getting
clusters
that
get
halfway
created
and
then
get
stuck,
and
we
can't
delete
them.
So
I
think
this
is
probably
the
sort
of
thing
where
we
need
to
take
it
on
a
case-by-case
basis,
make
a
best
effort
first
pass
through,
and
then,
if
there's
additional
things
that
the
crop
up,
we
can
do
them
in
follow
up
peers.
If
that
makes
sense.
B
We
have
a
nil
pointer
when
registering
an
instance
to
the
API
server,
ELB
and
I.
Read
this
one
this
morning
there
was
something
about
losing
some
subnet
IDs
because
of
a
bug
in
a
third
party
app
and
then
things
crashing
so
I'm,
not
sure
how
much
of
this
is
just
user
error
versus
stuff
that
we
can
protect
against,
but
it
probably
would
be
worthwhile
having
somebody
take
a
look
at
it.
B
B
Okay,
sorry
for
looking
that
I'll
deal
with
closing
that
and
or
at
least
getting
a
new
one
events
not
generated
when
create
release
disassociate
elastic
IP
fails.
This
is
one
that
I've
been
looking
at
with
Harish
and
there's
just
we
just
aren't
creating
events
for
these,
so
I
think
we
should
given
that
we
try
to
create
events
for
everything
else.
B
A
There
may
be
other
cases
where
we
create
like
a
transitive
dependency
like
when
we're
creating
the
route
tables.
We
may
not
generate
events
for
all
of
the
individual
operations
that
go
into
it.
We
may
just
be
generating
events
for
kind
of
the
high
level
resource
that
we're
creating,
so
it
would
probably
be
a
good
thing
to
do
a
quick,
walkthrough
and
just
verify
okay.
B
B
B
B
A
Yeah
I
think
looking
at
it
I
think
the
problem
is,
is
just
that:
we're
failing
to
do
a
check
for
empty
string.
Okay,.