►
From YouTube: CNCF Harbor's Community Zoom Meeting - Dec 1, 2021
Description
12/01/2021
[Stone Zhang] - Database Optimisation - Discuss issue https://github.com/goharbor/harbor/issues/15963
[Chenyu Zhang] - Demo https://github.com/goharbor/perf/wiki/How-to-setup-a-monitoring-environment
[Orlin Vasilev] - Peritus.io is added to the channels soon we can use it!
[Orlin Vasilev] - FOSDEM CFP opened "devroom containers"
[Orlin Vasilev] - Update on the GOVERNANECE.md - https://github.com/goharbor/community/pull/180
[Vadim Bauer] - https://github.com/szlabs/harbor-automation-4k8s - maybe Steven Zou can share thoughts on that repo
[Vadim Bauer] - https://github.com/container-registry/awesome-harbor
A
A
Please
add
your
topics
at
the
community
meetings
link
that
I've
provided
in
the
chat,
so
we
can
keep
track
of
what's
discussed
and
what
we
need
to
discuss
further
with
that.
I'm
gonna
share
my
screen
with
that
agenda
for
today
and
I'm
gonna
go
one
by
one,
so
stone
zhang,
I'm
sorry!
If
I'm
mispronouncing
names,
the
stage
is
yours:
do
you
want
to
share,
or
should
I
share
the
issue.
B
I
will
share
this
issue.
Okay,
okay,
let
me
share
my
screen.
C
B
Recent
weeks
we
are
working
on
improving
the
performance
of
the
internal
hardware
instance,
and
we
found
that
the
hardware
performance
is
heavily
relay
on
the
database.
Here
I
share
some
of
our
experience
in
a
hub
database
tuning
and
what
we
have
done
to
improve
it
first
is
the
idle
introduction
in,
however,
in
the
hub
instance,
with
hubble
2
1.3,
we
found
that
there
are
a
lot
of
database
connections
stay
in
idle
in
transaction
state
and
it
consumes
a
lot
of
database
connection
because
well
connection
in
the
stata
in
this
state.
B
B
B
After
harvard
2003,
we
have
skipped
most
of
the
read-only
request,
which
you
could
check
in
middlewares
dot
go
file,
we
have
middlewares.gov,
we
skipped
a
lot
of
read-only
requests.
B
This
is
the
first
one
I
need
to
share
and
the
second
one
is
the
is
related
to
the
hubble
installed
on
kubernetes,
when
we
want
to
scale
up
the
hard
core
part
for
a
better
performance.
B
Please
be
aware
that
the
max
database
connection
should
be
considered,
for
example,
when
the
max
open
connection
for
each
core
is
500,
and
there
is
also
have
a
limit
in
the
server
side.
This
is
the
database
server
and
the
server
side
limiting
the
maximum
connection
is
1500
when
they
have
a
core
product
in
three.
B
There
are
three
hardcore
products.
The
total
is
just
1500
or,
and
when
you
want
to
scale
up
the
product
to
four,
then
you
should
update
the
max
open
conduction
in
max
open
connection
to
300
for
each
con
for
each
hub
call
for
each
hub
core,
because
the
total
database
connection
max
total
database
connections
will
exceed
the
database.
B
The
max
database
database
connection
in
the
server
side
right.
If
we
we
just
scale
to
four,
then
we
are
2000
the
database
connections
and
it
will
exceed
the
database
settings
in
the
server
side.
B
One
day
this
connection,
you
can
see
the
server
side
that
we
are
the
dollar
client
that
will
receive
our
500
error
in
a
car
network.
Client
180,
excel.
B
B
B
And
another
topic:
we
need
to
talk
about
that.
A
lot
of
users
are
using
consequence
sql
with
master
slave
duplication
mode.
They
want
to
install
hardware
with
their
slave
instance,
but
so
far
hubble
doesn't
support
installation.
This
type
of
installation
there
are
some
middlewares
could
redirect
the
request
to
their
database
node
and
redirect
the
right
request
to
the
master
database
node.
B
If
you
have
any
experience
or
best
practice
on
this
topic
welcome
to
shared
in
our
computer
meeting
or
slack
channel.
Thank
you.
That's
all
for
the
database
improvement.
A
About
that,
do
you
think
that
that
can
be
formalized
in
some
some
form
of
blog
posts
or
something
with
with
some
measurement
and
to
be
published
is
like
a
white
paper
or
something
because
maybe
it's
not
just
you,
but
many
other
people
are,
maybe
they
have
the
same
the
same
problem
without
knowing
it
and
how
to
fix
that?
What
do
you
think.
B
Yeah
for
for
the
first
one
we
have
fixed
the
fix
the
e
fixed
in
2.3
and
the
further
yeah.
I
think
that
for
the
second
one
we
should
better
add
some
document
in.
B
But
currently
we
have
no
formal
document
for
this.
We
just
have
rpg.
D
A
D
That
would
be
great
yeah.
Also,
I'm
not
sure
vadim
is
on
the
call
right.
So
within
you
you,
you
know
you
you
previously,
you
mentioned,
you
have
run
some
ins
instance
in
regis
container
dot
com
right.
So
I'm
not
sure
if
you
face
any
performance
problem
in
I
mean
if
you
have
any
experience.
E
E
Yeah
we
had
performance
problems
specifically
with
the
database
connections
and,
as
you
already
mentioned
in
slide
number
two
that
that
it's
much
improved
in
in
in
version
2
2.3,
because
there
is
not
not
so
many
idle
connections
anymore
and
also
what
helped
for
us
is
kind
of
dropping
idle
connections
or
closing
having
you
know
this
time
out.
It
was
also
introduced
into
2.0
as
having
a
timeout
on
idle
connections.
E
So
now
that
the
connections
are
widely
reduced
and
also,
I
think
that,
on
the
slide
number
three,
you
mentioned
the
the
max
connection
limit,
but
okay,
you
have
to,
or
the
user
have
to
take
to
into
account
that
there
is
not
only
the
hardware
core
who
has
connected
to
the
database.
There
is
also
the
job
service
and
all
those
other
services,
so
they
all
accumulate
for
the
max
open
connection.
E
So
so
we
you,
as
a
user,
need
to
take
into
account
the
whole
state
or
the
whole
application
with
all
other
services
as
well,
because
they
all
contribute
to
the
to
the
connection
pool,
and
so
you.
A
E
Set
it
accordingly
and
not
from,
but
on
the
other
hand
I
mean
we
are
using
a
managed
postgres
database
on
our
site
so
and
it's
yeah
a
kind
of
big
one.
So
we
don't
see
that
much
of
the
problems
there
so
so
there's
from
from
our
site
is
it's
it's?
Okay,
it's
not
a
big
issue
except
the
connections,
and
now
it's
getting
better
with
the
connection.
So
the
connections
problem
is,
I
would
say,
it's
resolved
and
now
yeah.
E
Maybe
there
is
on
large
scale
query,
maybe
that
there
may
be
a
problem,
but
we
didn't
face
any
problems
so
far:
okay,
okay,.
D
Do
you
have
any
data
regarding
the
concurrent
poor
request
in
your
instance.
E
I
can
I
can
get
you
some
data,
if
you,
if
you,
if
you
want
yeah,
I
can,
I
can
ask
for
for
data,
and
what
do
you
have
something
particular
in
mind.
D
Yeah
we
want
to
know,
I
mean
just
scalability,
I
mean
how
many
can
kind
of
concurrent
pull
requests
normally
in
a
in
a
I
mean
a
real
world
hub
instance,
and
can
serve.
C
Okay,
so
we'd
also
like
to
have
what
they
are
post-request
that
face
connection
country.
I
mean,
if
you
can
share
the
graphic
about
the
database,
monitor
too
that
the
weak
one,
let's
see
if
there
are
any
log
in
the
database
and
if
you
want
you
can
share
the
the
trend
for
us
like
about
the
connection
call.
A
B
E
B
B
B
And
after
we
disabled
the
procurement
proton
and
the
the
calendar,
the
cost
can
be
four
thousand
key
for
four
thousand
yeah
frozen.
B
Yeah
we
have,
this
is
the
tesla
result
based
on
the
2.3.5.
E
Okay,
and
so
you
reduced
the
amount
of
locks
needed
for
from
400
000
to
400..
C
Yeah
in
the
for
4k
and
concurrency
pool,
we
noticed
that
there
are
solders
lock
in
the
database,
so
we
modified
some
code
for
the
for
update
of
updating
who
can't
at
all
time
so
that
patch
works
for
us
and
it
still
the
knockout.
C
But
by
the
way
our
change
is
dropping.
Another
proposal
to
have
a
foodie
solution
to
resolve
this
problem
in
2.5,
and
the
roughly
idea
is
that
the
we
want
to
use
async
mode
to
update
to
control
time
instead
of.
C
E
That's
that's
very
good.
C
F
Yeah
yeah,
okay,
hello,
everyone,
my
name
is
chang
zhang
and
I'm
from
real
vmware,
harvard
team,
and
today
I
want
to
share
about
how
to
set
up
a
monitoring
environment
with
harvard.
So,
firstly,
I
want
to
introduce
the
background
of
this
document.
F
For
example,
we
also
need
resource
usage
and
distributed
tracing
capabilities
to
help
locate
performance
bottlenecks.
So
so
here
is
an
example
to
show
how
to
set
up
a
monetary
environment
with
harbor,
okay,
so
first
we
need
to
deploy
harbor
and
we
can
download
offline
installer
to
install
harbor
this.
This
document
are
based
on
harvard
2
2
version,
because
the
trees,
the
trace
feature,
were
delivered
at
harbour.
2.4.
F
So
so
after
you
download
the
offline
installer,
you
need
to
change
your
hardware
configuration
there
are
two.
There
are
two
configuration
need
need
to
change.
One
is
you
need
anc
mount
metric,
related
fields
to
enable
hover
matrix,
because
the
default
value
is
false,
which
means
metric
wall
were
disabled
and
another
is
trace
trace.
You
need
an
amount
trace
rate
related
fields
to
enable
hover
tracing,
in
addition
to
set
the
set
tool
to
the
trees
enabled
another.
F
See
advisor
is
a
tool
which
can
analyze
resource
usage
and
the
performance
characteristics
of
running
containers.
It
is
a
open
source
project
from
google
and
the
post-grade
circle.
Exporter
can
expose
post-grade
circle,
database
metrics
and
the
promises
and
the
grand
final
and
jaeger
as
a
tool
which
has
the
ui
and
it's
usually
used
to
to
for
visualization
promisius,
is
a
systems
and
service
monitoring
system.
F
Jaeger
is
a
distributed
tracing
platform,
but
if
we
need
need
to
install
these
components
manually,
it
will
be
very
hard
and
it
will
cost
our
too
much
time
so
to
simplify
the
deployment
of
these
components.
We
have
integrated
a
simple
docker
composite
armor
for
one
click
deploy,
so
you
just
need
to
copy
below
two
yarmouth
files
to
the
hardware
install
founder.
F
F
C
advisor
harvard
db
exporter
from
sales,
grandparent
and
jaeger,
and
if
you
use
these
two
files
deployment
has
become
very
easy.
We
only
need
to
append
composer
and
operation
to
harbor
and
the
monitor
together
by
one
command.
F
You
just
need
append
one
monitor
dot
yammer
to
to
deploy
this
so
and
after
these
components
container
be
ready.
We
can
open
the
browser
with
this
address
to
access
graphene
and
this
for
jaeger,
but
for
grandfather
we
require
aiding
promises
as
data
cells
because
of
the
promises
is
running
in
the
same
network
with
grandfather.
F
Okay
and
after
you
do
the
you
do
these
steps,
you
can
do
some
operation
to
your
harbor
and
you
can
see
some
metrics
from
the
ground
now.
Okay,
so
let's
do
a
quick
demo
about
these
steps.
F
I
have
copied
these
two
files
to
my
environment.
One
is
from
sales,
armor
and
another
is
a
monitor.
C
Okay,
sorry,
you
have
to
share
your
terminal.
C
F
We
can
okay,
sorry
I
say
again:
I
have
copied
the
two
files
to
my
environment.
One
is
from
sales.yammer
and
another
is
a
monitor,
dot
yammer.
F
If,
if,
if
you
don't
change
any
other,
if
you
don't
custom
any
other
configuration
in
the
hardware
dot
cameras,
you
don't
need
to
change
anything
in
these
two
files.
So,
let's
deploy
just
by
one
command.
F
F
F
If,
if
you
first
install,
you
will
need
to
import
the
dashboards
manually,
it's
very
easy.
Just
click
the
sidebar
aid
button
and
use
import,
and
if
you
have
a
graphene
dashboard
id,
you
can
just
copy
it
and
load
it.
And
if
you,
if
your
configuration
was
provided
by
json,
you
just
passed
it
to
this
text
box
and
and
just
load
it
it
will
import
to
your
graph.
F
So
you
can
see
the
it
will
have
many
pen
panels,
grouped
by
grouped
by
some
some
some
some
future.
So
first
is
info,
it
represents
harbor
health
and
some
component
status,
and
some
these
are
some
general
metrics,
like
processor,
cpu
time
ghost
rates
and
guarantees
number
and
and
so
on,
and
next
is
the
core
matrix.
F
F
So
another
is
talker
c
advisor
dashboard.
This
will
have
two
parts.
One
is
note:
it's
a
summary
information
about
the
node
which,
like
how
many
containers
running
in
your
host
and
cpu
core
and
cpu
and
memory
usage
and
some
summaries
later,
okay,
and
by
the
all,
we
can
see
some
detail
of
details
on
some
containers
like
which
we
can
see
the
different
components:
different
container
use,
the
cpu
usage
and
memory
usage-
and
this
was
this-
is
a
disk
io
and
this
network.
F
We
need
to
change
the
net
interface
to
eth00
yeah
and
it
can
display
the
network
io
information,
okay,
the
last
one
is
postgraduate
card,
so,
firstly,
it
will.
It
has
some
generic
counters,
like
version
start
time,
max
connections
about
this
post-grade
circle
and
the
average
cpu
usage,
average
memory
usage
and
open
file
and
some
static
settings
about
the
possibly
circle
like
shared
buffers,
effective
catch
and
so
on
so
below
below
are
some
database
states
like
transactions,
log
tables
and
some
buffers
and
conflicts,
and
so
on.
D
F
A
And
again
I'll
laugh,
I
would
love
to
ask
you
if,
if
you
can
make
like
a
is
that
is
that
part
of
the
perf
repository
right
under
go
harbor.
A
Great
because
I
didn't
I
haven't
followed
that
repository,
so
that's
completely
new
for
me,
I'm
not
sure
how
much
of
that,
what
you
showed
is
new
and
or
it
was
existing
already,
but
that's
great
for
community
engagement,
I
mean
if
we
can
do
like
a
again
a
blog
post
or
a
short
video,
and
I
can
see
that's
only
local.
For
now
I
mean
you
use
docker
compose
to
run
that
locally.
A
A
A
And
yeah,
and
and
do
we
have
the
manifests
for
for
kubernetes,
or
we
have
them
only
for
the
ammo
files.
F
C
F
C
No
yeah:
let's
do
it
just
for
the
community
user
to
have
a
one-stop
setup
for
harvard
to
monitor
and
their
performance.
C
We
just
provided
a
guidance
for
the
community
unit
and
the
one
doctor
composed
file,
so
the
user
coming
from
the
community.
One
two
wants
to
have
contribute
on
the
kubernetes
version,
but
it
depends
on
his
schedule
there.
Okay.
A
A
Okay,
anyone
else,
because
my
my
my
topic
compared
to
the
previous
ones
are
is
just
ridiculous.
It's
I'm
not
sure.
If
you
can
recall,
we
had
a
demo
from
these
guys
from
prairie
test
for
the
chatbot
in
the
channel.
A
A
D
A
Another
topic
for
my
site
today:
all
of
them
opened
their
go
for
papers,
so
feel
free.
It's
going
to
be
virtual,
so
feel
free
to
submit
any
topics.
I
think
that
the
proper
the
room,
so
the
proper
track
will
be
container
containers.
A
And
don't
forget
to
submit
your
your
hopeful
papers
form
on
for
kubecon
eu
until
december
17th?
I
believe
I'm
going
to
put
this
to
over
here,
so
not
to
forget.
A
That's
a
continuation
of
this
one
and
two
updates.
Another
update
update
from
my
site.
There
is
a
pr
currently
going
for
the
the
governance
dock
needs
some
adjustment
to
implement
special
interest
groups,
and
that
would
be
sick
community,
which
will
for
now
will
contain
the
community
management,
the
program
management
and
the
documentation.
A
E
I
have
a
point
to
add
that
we
for
for
the
for
the
translations.
You
know,
I
think
it
would
make
sense
to
have
a
section
for
for
for
people
who
do
contribute
translations
for
harvard
you
know,
so
they
they
can
take
on
that
and
maintain
translations.
A
You
mean
we
can
create
a
work
group
for
this
one
and
like
a
sub
team
in
the
orc
and.
B
A
Yeah
yeah,
we
discussed
that
with
jonas
and
and
chris
azanczyk
from
cncf,
and
if
we
introduce
more
six
than
this
one
like
seek
docs
six
translations
seek
you
name
it
it
to
make
it
more
complicated
that
we
need
that
we
actually
need
it.
So
so
something
like
seek
community
engagement
is
good
enough
as
a
start
and
and
then,
if
we
need,
we
can
fork
out
the
sick
dogs,
for
example,
or
sick
translations.
If
the
effort
is
too
big
and
it's
like
a
different
governance
and
different
members,
okay
yeah
at
least
that's
my
my
understanding.
A
E
Yeah,
I
have
maybe
some
something
that
for
sharing
with
you,
I
created
a
a
new
project
which
is
called
you
know.
You
know
this
awesome
awesome
lists
that
are
there.
A
E
E
I
have
I
mean
if
we
have
still
time.
I
have
another
question
and
it's
about
the
the
the
project
that
steve
steven
is
working
on.
You
see
it
also
it's
a
project.
It's
number
two.
It's
called
a
harvard
project.
It's
called
harbor
automate,
4k
8s,
it's
a
part
of
the
day,
2
operations
and.
E
Maybe
I
would
like
to
ask
if
if
steven
could
tell
something
about
this
project,
I'm
just
pasting
it
here,
which
is
that's
a
very
interesting
project
and-
and
I
just
stumbled
upon
up
on
this
project
by
coincidence-
it's
not
it's
not
under
harbor.
It's
it's!
It's
it's
under
his.
I
see
that.
A
E
So
no
I'll
paste,
I
pasted
the.
E
E
D
Yeah
steven
you're-
not
here
today,
actually
he
is
joining
the
community
work
group.
I
mean
the
operator
or
group
to
work
on
some
tesco
day.
Two
operation-
maybe
there's
something
here
they
also
have
worked
on,
but
I'm
not
sure
about
that.
E
So
if
you
look
at
the
hardware
operator-
and
there
is
also
downstair
mentioned-
that
there's
a
reference
to
this
project-
and
I
just
stumbled
upon
by
coincidence
when
working
for
this
awesome
list-
and
this
looks
very
promising-
and
I
don't
know
if
we
could
kind
of
get
more
attention
on
that-
or
some
maybe
have
a
community
discussion
of
that
or
presentation
on
that
project
in
the
community.
This
would
be,
I
think,
very
helpful
to
a
lot
of
people.