signoz / signoz Goto Github PK

While flattening the data, we should also store Events data from spans to Apache Druid.
This contains exception data apart from reasons for erroneous spans like Redis timeout

Currently Error Processing field is done by:

• Presence of tag error:true
• http.status_code tag with value >500

Need to add another logic:
Span contains a field Status which works for all sorts of client & server requests.

Suggesting, testing of conditions to check whether the present 2 conditions are needed or can be done alone by Status field in Span.

As an open source project it is weird to see posthog analytics tracking everywhere.

https://github.com/open-telemetry/opentelemetry-collector/releases/tag/v0.19.0

In the above image, /traces endpoint is loaded but Metrrics is selected in the left panel.

Once we select a service from the dropdown in the service map, the dropdown does not give the option to show any other default view to deselect the service.

Add an option in the service-picker dropdown to choose a default view without any service selected.

Could you look to http://clickhouse.yandex/docs/en/?
Clickhouse is a blazing fast opensource (Apache 2.0) columnar database which have near-linear scalability
and it successfully applied as a Jaerger storage jaegertracing/jaeger#1438

Maybe it will better use this database as storage backend for your use-case?

Currently, to filter spans by tags there are 2 operators:

• EQUALS
• CONTAINS

An operator with the capability to check for IS NOT NULL values will be useful in getting all those spans which have a value of the tag. The value of IS NOT NULL operator that is passed in the API call would be isnotnull

Change title to DataBase Calls Avg Duration (in ms)

• Add a filter to identify pii/sensitive information in trace data and different metrics related to it.
• Flagging or encrypting request which has PII or sensitive data, so enterprises can monitor all the sensitive data in transit.

Hi team
I've tried out SigNoz with Spring boot micro-services which are already integrated with Zipkin. In the dashboard, I can see the /webjars and /swagger endpoints. It would be great to have some 'skip pattern' to avoid un-desired URLs. Thanks!

Change title from Errors per sec to Error Percentage (%) and same in legends beside the red circle

The backend API has all parent-child relationships with numCounts assigned to each edge which is not needed now. Maybe we can add that info later to any relevant UI component. Below are functionalities to be met:

• Size of Nodes according to the RPS being handled by the application

• Choosing a service from dropdown should bring that node to the center of the page

• On hover/select on node show popup with info from API /api/v1/services?start=xxx&end=xxx

• RPS, error percent, and p99 latency should be displayed on hover/select

• Color logic of node -> if errors then Red, if 4xx then Yellow else Green

Checkout lib https://github.com/crubier/react-graph-vis, which uses parent lib https://github.com/visjs/vis-network and examples at https://visjs.github.io/vis-network/examples/

https://github.com/Netflix/vizceral → another library that has dots on the arrow between nodes. Though this library does not seem to be maintained anymore.

This product also has moving data lines → https://app.lightstep.com/play/explorer?snapshot_id=HrPQVHjZX0&active_tab=service-diagram&selected_node_id=krakend-api-gateway

We need to show some type of movement in lines to depict the direction of data movement between nodes like an arrow. Makes the chart feel dynamic rather than static.

How can we enable easy instrumentation of Go applications

Currently, the SpanAggregation has startTime and endTime equal to the traces page which might be less relevant if the period is short. For now, we can keep a default 30 mins window for SpanAggregation API call

Currently, the website and documentation of SigNoz are hosted in private repos. We should plan to shift documentation to this public repo so that other contributors building and testing different integrations, deployment methods, and changes to docs can be democratically and publically updated.

We use Docusaurus for documentation and will post the steps to shift soon.

Currently, it works based on the time range chosen from the DateTime picker widget at the top-right of the page. We need to make Service Map just show the last 1 min data from the API even for the API call to the /services endpoint.

This will make ServiceMap work like a snapshot of architecture at the current moment.

Requirements:

• Remove the custom DateTime picker option
• Keep the Refresh button at the top-right as it is. It should just update the data and chart
• Change the services API call to fetch just the last 1-minute data.

It works by single clicking at the datapoints in the chart. It won't work for full plotted lines but only where the timestamp has data. Currently, the resolution of data is 1 min in the chart. Also, you would see small circles on hovering at presence of datapoint as shown in image attached.

Suggesting to add "Span Kind" filter in top panel of span filtering aside of Latency Filter in the below screenshot.

The permissible values of Span Kind would be:

1. SERVER which shall internally transform to kind=2 in the API call
2. CLIENT which shall internally transform to kind=3 in the API call

Hi team
We've many Spring boot micro-services and under metrics tab all those micro-services are visible. On click of it a service (name), we can see various dashboards like latency, request per sec, errors etc., for the 'selected' service.

Is there any feature by which we can view an 'All In One' Dashboard where all the micro-services are viewable may be with specific type of dashboard (req/s or latency). Would love to have it :D

Thanks!

Internally timestamps from frontend gets converted to format startTime.Format("2006-01-02 15:04:05") which may lead to difference in timezones.

As we are trying to make SigNoz deployment to work in common infra setups, a good first issue could be just trying installing SigNoz in their local setup and sharing if they face any issue.

Please share your infra details and did you face any issue in running SigNoz

Providing a single click Heroku deploy option may be a good idea to make it easy to test SigNoz for developers.

Tasks

• Check if current SigNoz docker can run in Heroku free tier and the compute requirements are served in the free tier of Heroku. Goal is not to test with high load but to get the application up and running with sample Hot ROD application

Suggest an alternate view along with flamegraph?

Should the spans with errors be coloured RED in flamegraph?

Need to fix the zoom/focus to node in wider screens when a service is selected from the dropdown service picker

Currently, if we add more than 10 chars to the service name in the node, the text overflows the node. Need to find a way to allow up to 20 chars as service name inside the node.

Current Behaviour

• Database calls tab is shown even when there is no data in the graphs ( i.e. no database calls were made)
• The graphs in the database tab are empty

Expected Behavior

• Database tab shouldn't be shown for a service for which there are no DB calls made

Publish typical benchmarks and calculations on what would it cost to self host SigNoz at different scale and sampling.

@pranay01 can you share content to be added

To have long term data storage, s3 should be setup. This is not currently clear from the docs

Current Behaviour
SignUp page after install shows old logo

Expected Behaviour:
Signup page should show the updated logo

The time displayed in spans list and aggregate metrics on Traces page are not in the same time range.

Steps to reproduce

• Open Application detail page by clicking a service
• Reduce browser width size
• After some width size, the API endpoint table overflows

Steps to reproduce

1. Install the application
2. Switch to light mode using the toggle at the top
3. Got to Add instrumentation page by navigating through the left panel
4. The page shows dark background

Expected Behavior

Dark background in this page shouldn't be shown

Keeping the startTime and endTime as the same would be more beneficial to debug issues. As there can be 100s of spans in a sec and thus a time range of 15 mins does not make much sense in debugging in the trace page.

@pranay01 what do you think?