I tried uploading a Ditto generated TD in a github gist, but it can't be parsed the right way. Looking at their implementation i think maybe it could be a issue with the fact that Ditto generates "nested" properties (Features in another TD).
I tried uploading a Ditto generated TD in a github gist, but it can't be parsed the right way. Looking at their implementation i think maybe it could be a issue with the fact that Ditto generates "nested" properties (Features in another TD).
speaked to the node-wot team, and i'm raising a github issue.
2 replies
What is the connection configuration to be able to map command responses?
For modify command we listen to twin events for change propagation.
@tobias-zeptio the configured "payloadMapper" of the "Source" should also be used for command responses - there is currently no way to configure another payload mapper for the responses differently
the
the
"outgoingScript" of the configured JS payload mapper is then executed for the command responses (which also are Ditto Protocol messages, so the signature to implement is the same)
(In the Javascript Mapper)
Hi @thjaeckle , I imported ditto into my eclipse workspace to go through the code, and was facing certain dependency issues (image1: https://imgur.com/F6zWgAT, image 2:https://imgur.com/DyvEsZx) Do I manually get the plugins and add them to respective pom.xml's? I do not want to break the build therefore I wanted to know if that is right.
4 replies
The plugin somewhat helped, but it still has a lot of errors which cannot be explained since the mvn clean install and the ditto build in general runs flawlessly on my machine. It's just the IDE that gives error. Should I try to resolve them(if I am to contribute I might need to test the code).
Hi ppl,
Firstly I'd like to thank all the hard work put into ditto, it's a pretty amazing piece of software!
Said that, I'm currently trying to do some load test on it, here where I work we intend to use it to manage around 300k devices and would like to have something around 6k~7k device sincronizations/s. We use k8s currently and I'm doing all my tests using helm chart for ditto and percona mongodb.
I'm having a bit of strugle to achieve 1k msgs/s. My test is basically create 100k devices and start something around 30 pods sending around 500 msgs/s.
Currently I'm using the following configurations:
nginx: 15 pods, cpu: 80m, memory: 10Mi
Gateway: 6 pods, cpu: 800m, memory: 4000Mi
concierge: 7 pods, cpu: 600m, memory: 4000Mi
policies: 1 pods, cpu: 400m, memory: 4000Mi
things: 5 pods, cpu: 800m, memory: 4000Mi
thingsSearch: 10 pods, cpu: 800m, memory: 4250Mi
connectivity: 5 pods, cpu: 1000m, memory: 4000Mi
Would anyone have some idea on were I could increase the resourses, or maybe an article or documentation to help me to make this decision in order to achieve the desired throughput?
Firstly I'd like to thank all the hard work put into ditto, it's a pretty amazing piece of software!
Said that, I'm currently trying to do some load test on it, here where I work we intend to use it to manage around 300k devices and would like to have something around 6k~7k device sincronizations/s. We use k8s currently and I'm doing all my tests using helm chart for ditto and percona mongodb.
I'm having a bit of strugle to achieve 1k msgs/s. My test is basically create 100k devices and start something around 30 pods sending around 500 msgs/s.
Currently I'm using the following configurations:
nginx: 15 pods, cpu: 80m, memory: 10Mi
Gateway: 6 pods, cpu: 800m, memory: 4000Mi
concierge: 7 pods, cpu: 600m, memory: 4000Mi
policies: 1 pods, cpu: 400m, memory: 4000Mi
things: 5 pods, cpu: 800m, memory: 4000Mi
thingsSearch: 10 pods, cpu: 800m, memory: 4250Mi
connectivity: 5 pods, cpu: 1000m, memory: 4000Mi
Would anyone have some idea on were I could increase the resourses, or maybe an article or documentation to help me to make this decision in order to achieve the desired throughput?
Do you have insights via metrics via Prometheus and Grafana? That would be the easiest way to detect bottlenecks..
I noticed you scaled the search the most. We are currently working on a optimized search index which will drastically improve performance. The current search index is not optimized for complex thing structured and puts the Mongodb under a lot of CPU pressure.
@thjaeckle thanks for the insights,
So I'm testing sending the messages through the /cloudevents endpoint.
My idea is to publish the events into a rabbitmq using the connectivity.
I increased the search because I saw that it was one of the modules that more throttled. Right now my tests only change a value for a prop on a specific feature, suing the topic "things/twin/commands/modify". It's not that complex.
I have some metrics for k8s resources and found out some metrics from each pod as well, I'm trying to guide my adjusts through this metrics either.
I'll try to increase the cpu for the modules and see if I'm able to achieve a higher throughput. I'll let you know the results thanks :)
So I'm testing sending the messages through the /cloudevents endpoint.
My idea is to publish the events into a rabbitmq using the connectivity.
I increased the search because I saw that it was one of the modules that more throttled. Right now my tests only change a value for a prop on a specific feature, suing the topic "things/twin/commands/modify". It's not that complex.
I have some metrics for k8s resources and found out some metrics from each pod as well, I'm trying to guide my adjusts through this metrics either.
I'll try to increase the cpu for the modules and see if I'm able to achieve a higher throughput. I'll let you know the results thanks :)
but @aleshark87 you are referring to a Gradle project, which uses Eclipse Ditto … which is different to working on the Eclipse Ditto code itself, which doesn't use Gradle, right?
Now i'm trying to set that up for receiving Messages, but i can't get any Event. I have a webapp(with sse working for properties update) and i'm trying to send messages via the ditto api page.
new EventSource('http://localhost:8080/api/2/things/com.project.thesis:greenhouse01/outbox/messages/subject', {withCredentials: true}).onmessage =
function (event){
if(event.data != ''){
console.log(event);
}
console.log(event);
//$('#eventParagraph').text(event);
}
and this is the curl request that the ditto api does:
curl -X 'POST' \
'http://localhost:8080/api/2/things/com.project.thesis%3Agreenhouse01/outbox/messages/subject?timeout=0' \
-H 'accept: */*' \
-H 'Authorization: Basic ZGl0dG86ZGl0dG8=' \
-H 'Content-Type: application/json' \
-d '"test"'
I tried also creating a
EventSource attached to http://localhost:8080/api/2/things where, if i understand the ditto SSE mechanism, i should receive all events. I'm quite sure that the way i send the messages is right, since i receive that via WebSocket. There is something i can do for debugging ?
3 replies
Would you be available sometime next week to discuss the current state (conflicts with main branch)?
2 replies
I was expecting the merge of my PR eclipse/packages#371 to generate the release of a new chart version, but the latest chart version is still 2.5.0 in helm. Is the release a manual process, or what remains to be done?
3 replies
if your "as a user" has access to the database - yes
Ditto uses Event sourcing which means that event (change) is persisted to the DB ..
by default Ditto deletes stale events in the background, but this can be deactivated via configuration
Ditto uses Event sourcing which means that event (change) is persisted to the DB ..
by default Ditto deletes stale events in the background, but this can be deactivated via configuration
My goal is to create a tiny IoT AWS. I m using a cloud2edge cluster (locally for now) and i want to stay as close as possible to the functions of AWS.