2019-12-19 09:25:54 UTC - Ning yougang: Hi, guys, seems current sharding pool
balancer get managed invokers and blackbox invokers regardless of their healthy
stats
I am not sure it is a bug or not, it seems before calculate managed/blackbox
invokers, should based on `healthy invokers` only.
+1 : Dominic Kim, Jiang PengCheng
https://openwhisk-team.slack.com/archives/C3TPCAQG1/p1576747554005700
----
2019-12-19 09:26:24 UTC - Ning yougang: Ho do you guys think?
https://openwhisk-team.slack.com/archives/C3TPCAQG1/p1576747584006000
----
2019-12-19 09:31:37 UTC - Markus Thömmes: If I'm not mistaken it should already
take that into account, but it's been a while :smile:
https://openwhisk-team.slack.com/archives/C3TPCAQG1/p1576747897006500
----
2019-12-19 09:33:04 UTC - Markus Thömmes: yeah:
```if (invoker.status.isUsable &&
dispatched(invoker.id.toInt).tryAcquireConcurrent(fqn, maxConcurrent, slots)) {
Some(invoker.id, false)
}```
https://openwhisk-team.slack.com/archives/C3TPCAQG1/p1576747984006700
----
2019-12-19 09:33:16 UTC - Markus Thömmes: it only picks usable invokers
https://openwhisk-team.slack.com/archives/C3TPCAQG1/p1576747996007000
----
2019-12-19 09:34:04 UTC - Ning yougang: Oh, let me check the latest codes ^^
https://openwhisk-team.slack.com/archives/C3TPCAQG1/p1576748044007400
----
2019-12-19 09:34:33 UTC - Markus Thömmes: this is from the `schedule` function
https://openwhisk-team.slack.com/archives/C3TPCAQG1/p1576748073007700
----
2019-12-19 09:41:23 UTC - Dominic Kim: It seems managed / blackbox invokers to
use is defined without taking health states into account?
https://openwhisk-team.slack.com/archives/C3TPCAQG1/p1576748483008900
----
2019-12-19 09:42:10 UTC - Dominic Kim: So if there is no healthy blackbox
invokers, it will end up with `No invokers available` even if there are still
resources in managed invokers.
https://openwhisk-team.slack.com/archives/C3TPCAQG1/p1576748530009800
----
2019-12-19 09:42:30 UTC - Dominic Kim: But not quite sure it is desired to use
managed invokers if there is no available blackbox invoker.
https://openwhisk-team.slack.com/archives/C3TPCAQG1/p1576748550010400
----
2019-12-19 09:42:31 UTC - Markus Thömmes: right, that's by design I think
https://openwhisk-team.slack.com/archives/C3TPCAQG1/p1576748551010500
----
2019-12-19 09:43:09 UTC - Markus Thömmes: ah I see, the "bug" is that the list
of managed and blackbox containers is not calculated based on actual health?
https://openwhisk-team.slack.com/archives/C3TPCAQG1/p1576748589011000
----
2019-12-19 09:43:19 UTC - Dominic Kim: Yes
https://openwhisk-team.slack.com/archives/C3TPCAQG1/p1576748599011200
----
2019-12-19 09:43:48 UTC - Markus Thömmes: I think that's by design in that
blackbox containers are only seperated because they have been deemed more
"dangerous" to the underlying machines because you can willynilly do whatever
you want in them
https://openwhisk-team.slack.com/archives/C3TPCAQG1/p1576748628011900
----
2019-12-19 09:44:10 UTC - Markus Thömmes: if you managed to abuse that and we'd
recalculate the list based on unhealthiness, an attacker could still get to all
invoker machines eventually
https://openwhisk-team.slack.com/archives/C3TPCAQG1/p1576748650012400
----
2019-12-19 09:44:40 UTC - Markus Thömmes: now I think it's very debatable if
this separation is really necessary, but I think ^^ is the reasoning for the
current behavior
+1 : Ning yougang, Bill Zong
https://openwhisk-team.slack.com/archives/C3TPCAQG1/p1576748680013000
----
2019-12-19 09:45:34 UTC - Dominic Kim: Agree. I wondered the separation is
really required.
https://openwhisk-team.slack.com/archives/C3TPCAQG1/p1576748734013400
----
2019-12-19 13:40:16 UTC - Dave Grove: We should consider changing the defaults
to not separate the two kinds of invokers. Still support the option, but not
make it the default. Having the split invoker pools seems to cause a fair
amount of user confusion for people starting off with the system
https://openwhisk-team.slack.com/archives/C3TPCAQG1/p1576762816015200
----
2019-12-19 13:47:15 UTC - Markus Thömmes: I'd also suggest we should discuss if
and why we still need it
https://openwhisk-team.slack.com/archives/C3TPCAQG1/p1576763235015600
----
2019-12-19 13:47:43 UTC - Markus Thömmes: the general security concern is a
little obscure to me, another concern is that downloading images of arbitrary
size can fill out disks
https://openwhisk-team.slack.com/archives/C3TPCAQG1/p1576763263016200
----
2019-12-19 19:14:36 UTC - giusdp: Hey there, what would the best way to run a
kubernetes cluster having my computer as the master node and a remote virtual
machine as a worker node, to then deploy openwhisk on it? I made a cluster with
kubeadm and weavenet for networking but im having some problems with the
deployment. Maybe I'm missing some things i should know about kubernetes... ?
https://openwhisk-team.slack.com/archives/C3TPCAQG1/p1576782876019300
----
