2021-01-22 16:03:48 UTC - Mina Michael: @Rodric Rabbah Yes. Attach metadata to 
action and make it visible to the scheduler. The scheduler is able to pull out 
information about a pod that needs to be scheduled. I'm trying to add to this 
information.
https://openwhisk-team.slack.com/archives/C3TPCAQG1/p1611331428019400?thread_ts=1611260141.005400&cid=C3TPCAQG1
----
2021-01-22 16:05:42 UTC - Mina Michael: @Rodric Rabbah About the property that 
specifies a scheduler for the pod, no I didn't figure it out. However, I 
disabled the default scheduler, and now our schedule can just schedule 
everything.

It would be better for the scheduler to schedule just the pods that we need. 
Specifying the "schedulerName" property would be, I guess, a cleaner way of 
doing things.
+1 : Rodric Rabbah
https://openwhisk-team.slack.com/archives/C3TPCAQG1/p1611331542019600?thread_ts=1611260141.005400&cid=C3TPCAQG1
----
2021-01-22 16:09:42 UTC - Rodric Rabbah: I think using an `annotation` on the 
action is the likely best implementation approach for you, the annotation is 
not available inside the action container but is available in the invoker.
https://openwhisk-team.slack.com/archives/C3TPCAQG1/p1611331782019900?thread_ts=1611260141.005400&cid=C3TPCAQG1
----
2021-01-22 16:15:40 UTC - Mina Michael: @Rodric Rabbah You mean, the invoker 
pod? ...when a pod needs to be scheduled, I get information about that pod. The 
invoker pod (if that's what you mean) is not really involved in the process. 
I'm not sure how to get the information from the invoker pod.
https://openwhisk-team.slack.com/archives/C3TPCAQG1/p1611332140020100?thread_ts=1611260141.005400&cid=C3TPCAQG1
----
2021-01-22 16:29:28 UTC - Mina Michael: @Rodric Rabbah Like I said, I'm trying 
to get metadata about the function. An alternative solution would be to get the 
actual code of the function. Do you think that's possible?
https://openwhisk-team.slack.com/archives/C3TPCAQG1/p1611332968020300?thread_ts=1611260141.005400&cid=C3TPCAQG1
----
2021-01-22 16:38:39 UTC - Rodric Rabbah: either is plausible - if you GET the 
action using the API or couchdb directly you have access to the metadata (and 
annotations) and/or the code.
https://openwhisk-team.slack.com/archives/C3TPCAQG1/p1611333519020500?thread_ts=1611260141.005400&cid=C3TPCAQG1
----
2021-01-22 16:45:14 UTC - Mina Michael: @Rodric Rabbah you mean, like, use the 
wsk client to get the code or the annotations?
https://openwhisk-team.slack.com/archives/C3TPCAQG1/p1611333914020700?thread_ts=1611260141.005400&cid=C3TPCAQG1
----
2021-01-22 16:50:57 UTC - Rodric Rabbah: The wsk cli is among an HTTP request 
to the api host. You can use the cli or call the API directly. If you use wsk 
-d you can see the api calls.  
https://openwhisk-team.slack.com/archives/C3TPCAQG1/p1611334257022300?thread_ts=1611260141.005400&cid=C3TPCAQG1
----
2021-01-22 16:53:06 UTC - Mina Michael: @Rodric Rabbah Okay, I'll try to get 
the HTTP request from the wsk cli.

But is there way to directly get the function code or annotations without 
having to use HTTP? I mean, it's a bit strange to have the master do an HTTP 
request to itself, right?
https://openwhisk-team.slack.com/archives/C3TPCAQG1/p1611334386022500?thread_ts=1611260141.005400&cid=C3TPCAQG1
----
2021-01-22 16:56:25 UTC - Rodric Rabbah: i don’t understand your setup really - 
at some point the action is fetched and resources scheduled. In openwhisk’s 
standard control plane, this is done by the `invoker` which will fetch the 
action (metadata and code) and allocates resources accordingly.
https://openwhisk-team.slack.com/archives/C3TPCAQG1/p1611334585022700?thread_ts=1611260141.005400&cid=C3TPCAQG1
----
2021-01-22 17:01:01 UTC - Mina Michael: @Rodric Rabbah I just have openwhisk 
installed over Kubernetes. I used helm to do that. I've replaced the 
Kubernetes' default scheduler with a custom one. I've removed the pod 
responsible for scheduling, and I run a python script that does the scheduling 
instead of the pod.

For now, we've said that this python script can get information about the pod 
that it needs to schedule by making an HTTP request. (The script runs on the 
master node).
https://openwhisk-team.slack.com/archives/C3TPCAQG1/p1611334861022900?thread_ts=1611260141.005400&cid=C3TPCAQG1
----
2021-01-22 19:18:29 UTC - Brendan Doyle: ^ just to add it's a limit you need to 
configure so that you don't let your machine run more concurrently than your 
ram can handle. There's always overhead and / or other things on the machine 
using ram as well. Tuning this number is an art to get just right
https://openwhisk-team.slack.com/archives/C3TPCAQG1/p1611343109023200?thread_ts=1611269277.018900&cid=C3TPCAQG1
----
2021-01-22 21:08:13 UTC - Alessandro Banfi: I changed the settings and I'm now 
finally able to run more than 4 actions! Thank you very much for your advice!
partyparrot : Rodric Rabbah
https://openwhisk-team.slack.com/archives/C3TPCAQG1/p1611349693023400?thread_ts=1611269277.018900&cid=C3TPCAQG1
----

Reply via email to