.changelog-archive/CHANGELOG.08.md
Upload file button now opens the modal correctly. #48766, @ashharrison90For a data source query made via /api/ds/query :
DatasourceQueryMultiStatus feature is enabled and
DataResponse, the resulting HTTP status code is now 207 Multi Status instead of 400 Bad gatewayDatasourceQueryMultiStatus feature is not enabled and
DataResponse, the resulting HTTP status code is 400 Bad Request (no breaking change)
--> Issue #48550/api/v1/rules. #45969, @gotjoshFor a proxied request, e.g. Grafana's datasource or plugin proxy:
499 Client closed request instead of 502 Bad gateway504 Gateway timeout instead of 502 Bad gateway. Issue #47473The change in behavior is that negative-valued series are now stacked downwards from 0 (in their own stacks), rather than downwards from the top of the positive stacks. We now automatically group stacks by Draw style, Line interpolation, and Bar alignment, making it impossible to stack bars on top of lines, or smooth lines on top of stepped lines. Issue #47373
The meaning of the default data source has now changed from being a persisted property in a panel. Before when you selected the default data source for a panel and later changed the default data source to another data source it would change all panels who were configured to use the default data source. From now on the default data source is just the default for new panels and changing the default will not impact any currently saved dashboards. Issue #45132
The Tooltip component provided by @grafana/ui is no longer automatically interactive (that is you can hover onto it and click a link or select text). It will from now on by default close automatically when you mouse out from the trigger element. To make tooltips behave like before set the new interactive property to true.
Issue #45053
/api/tsdb/query API has been deprecated and will be removed in a future release. Use /api/ds/query instead. Issue #45238
When user is using Github OAuth, GitHub login is showed as both Grafana login and name. Now the GitHub name is showed as Grafana name, and GitHub login is showed as Grafana Login. Issue #45438
The meaning of the default data source has now changed from being a persisted property in a panel. Before when you selected the default data source for a panel and later changed the default data source to another data source it would change all panels who were configured to use the default data source. From now on the default data source is just the default for new panels and changing the default will not impact any currently saved dashboards. Issue #45132
<!-- 8.4.11 START -->AngularJS plugin support is now in a deprecated state, meaning it will be removed in a future release. Currently, that is planned for version 10 (in 2023). The documentation site has an article with more details on why, when, and how. Issue #45149
<!-- 8.4.0 END --> <!-- 8.4.0-beta1 START -->orgs roles, removed fixed:orgs:reader introduced in beta1. #42049, @gamabStarting with Grafana v8.3.0, if you have not explicitly disabled unified alerting and do not have legacy alerts set up you are automatically "migrated" to Grafana 8 Alerting.
A migration from legacy to Grafana 8 Alerting will never incur a data loss, as the previous data is kept around for rollback purposes. However, going from Grafana 8 Alerting to legacy alerting will delete all the data created for Grafana 8 Alerting. It is recommended that you backup your database before attempting a migration between systems.
If unclear, please verify the table below:
[alerting][enabled] | [unified_alerting][enabled] | With Existing Legacy Alerts | Result |
|---|---|---|---|
true | true | N/A | Error |
true | false | N/A | Legacy Alerting |
true | not set | Yes | Legacy Alerting |
true | not set | No | Grafana 8 Alerting |
| not set | true | N/A | Grafana 8 Alerting |
| not set | false | N/A | Legacy Alerting |
| not set | not set | Yes | Legacy Alerting |
| not set | not set | No | Grafana 8 Alerting |
false | true | N/A | Grafana 8 Alerting |
false | false | N/A | Alerting disabled |
false | not set | N/A | Grafana 8 Alerting |
N/A in the "With Existing Legacy Alerts" column means that it does not matter if you have legacy alerts or not. Issue #42200
In Grafana 8.3.0-beta2 we changed how alert rules that use Keep Last State for If execution error or timeout are upgraded from Legacy Alerting to Grafana 8 alerting. In 8.3.0-beta1 and earlier, alert rules with Keep Last State for If execution error or timeout were changed to Alerting when upgrading from Legacy Alerting to Grafana 8 alerting. However, in 8.3.0-beta2 these alert rules are now upgraded to a new option called Error. With this option, on encountering an error evaluating an alert rule, Grafana creates a special alert called DatasourceError with the rule_uid and ref_id as labels and an annotation called Error with the error message. Issue #41869
The access mode "browser" is deprecated in the following data sources and will be removed in a later release:
In Grafana 8.2.5 and later, this change fixes a bug in the evaluation of alert rules when using the AND operator to compare two or more conditions. In Grafana 8.2.4 and earlier such alert rules would evaluate to OK if at least one, but not all, conditions returned no data. This change fixes that bug such that in Grafana 8.2.5 these alert rules now evaluate to No Data.
If an alert should evaluate to OK when one or all conditions return No Data then this can be done via changing If no data or all values are null to OK. However, this will not preserve the old behaviour in 8.2.4 where an alert will be OK if at least one, but not all, conditions return no data and then No Data if all conditions return No Data. Issue #41305
grafana-server service from starting on older versions of systemd, present on Ubuntu 18.04 and slightly older versions of Debian. If running one of those versions, please wait until v8.2.1 is released before upgrading. If you still want to upgrade or have already ugpraded, a simple fix is available here: https://github.com/grafana/grafana/issues/40162#issuecomment-938060240 Issue #38109ngalert feature toggle and introduce two new settings for enabling Grafana 8 alerts and disabling them for specific organisations. #38746, @papagianGrafana v8.2.0-beta1 caused data loss for users having enabled ngalert in 8.0.x - 8.1.x and created alerts using the new alerting system. This issue is now fixed except if the deployment has multiple organizations and the feature flag was enabled in the previous version (8.0.x - 8.1.x).
In this scenario (upgrade from 8.0.x - 8.1.x with multiple organizations and ngalert enabled to 8.2.0-beta2), the migration will assign existing notification policies and contact points to the first organization and then apply the default alertmanager configuration to all organizations. This will effectively reset notification policies for all organizations. Issue #39541
ngalert feature toggle it has been deprecated it will be removed in a future release. To enable Grafana 8 alerts, modify your configuration and:
unified_alerting section set the enabled property to truealerting section set the enabled property to false Issue #38746org instead of user. #39353, @gotjoshresponse_limit for datasource responses. #38962, @dsotirakistrace to logs feature. #38943, @connorlindsey/var/run to /run. #35739, @MichaIngThe monaco-editor dependency in grafana-ui has been updated to a newer version (0.27.0), which is not completely backward compatible with the old version (0.21.2). The backward incompatible changes are fairly small, but they do exist, so if your code accesses the raw monaco-objects through the grafana-ui package, please check the monaco-editor changelog and apply any necessary changes. Issue #39027
The mandatory css prop in grafana/ui components has been removed.
Previous versions of grafana/ui components were typed incorrectly due to a dependency mismatch between emotion 10 and 11 causing a css prop to be added to components that extended react types.
Issue #38078
Grafana v8.2 fixed an issue with org isolation for notification configuration but to fix this Grafana will now re-run the migration from old alerting and this will cause complete removal of all new alert rules and notification configurations. This data loss is not something we find acceptable and are working on ways to mitigate it. So in the meantime, if you are an early adopter of unified alerting please wait with trying v8.2 beta. Issue #37414
Panel queries and/or annotation queries that used more than one statistic will be converted into one query/annotation per statistic. In case an alerting rule was based on a query row that had more than one statistic, it would now be based only on the first statistic for that query row. New alerting rules will not be created for migrated queries. Please note that in most cases it would not make sense to have an alerting rule that is based on multiple statistics anyway. Issue #36925
getHighlighterExpressions in datasource APIs ( used to highlight logs while editing queries) has been deprecated and will be removed in a future release.
ExploreQueryFieldProps interface for query editors has been deprecated and will be removed in a future release. Use QueryEditorProps instead. Issue #38942
/ as separator. #39185, @dsotirakisfalse into true. #37402, @simPoddataproxy_max_idle_connections config variable. #35864, @dsotirakisWhen parsing Elasticsearch query responses using template variables, each field gets named after the variable value instead of the name.
For example, executing a terms aggregation on a variable named $groupBy that has @hostname as a value, the resulting column in the table response will be @hostname instead of $groupBy Issue #36035
Azure Monitor data source no longer supports different credentials for Metrics and Logs in existing data sources. To use different credentials for Azure Monitor Logs, create another data source. Issue #35121
Existing Azure Metrics Logs queries for Log Analytics Workspaces should be backward compatible with this change and should not get impacted. Panels will be migrated to use the new resource-centric backend when you first edit and save them.
Application Insights and Insights Analytics queries are now read-only and cannot be modified. To update Application Insights queries, users can manually recreate them as Metrics queries, and Insights Analytics are recreated with Logs.
Issue #33879
fieldConfig when changing the panel type. #35676, @ashharrison90The following endpoints were deprecated for Grafana v5.0 and support for them has now been removed:
/dashboards/db/:slug/dashboard-solo/db/:slug/api/dashboard/db/:slug/api/dashboards/db/:slug Issue #35104The default HTTP method for Prometheus data source is now POST. Previously, it was GET. The POST APIs have been available since January 2018 (Prometheus 2.1.0) and they have fewer limitations than the GET APIs. For example, when dealing with high cardinality labels, GET hits the URL size limit.
If you have a Prometheus instance with version < 2.1.0, which uses the default HTTP method, update your HTTP method to GET. Issue #34599
<!-- 8.0.0-beta3 END --> <!-- 8.0.0-beta2 START -->selected states on dashboard load. #34197, @hugohaggmarkThe workspaces() template variable, mainly for use with Azure Monitor Logs, has been changed to return resource URIs instead of Log Analytics Workspaces GUIDs. This should not impact Azure Monitor Logs queries, but if the variables are being used in other data sources which expect a Workspace GUID may no longer be compatible.
Custom template variables used in the workspace or resource field in Azure Monitor Logs queries should resolve to an Azure Resource URI in the format /subscriptions/{guid}/resourceGroups/{resource-group-name}/{resource-provider-namespace}/{resource-type}/{resource-name}
Issue #34445
Removes support for deprecated backend plugin protocol (v1) including usage of github.com/grafana/grafana-plugin-model.
Issue #34127
grafana. #33925, @bergquistRemoves the never refresh option for Query variables. Existing variables will be migrated and any stored options will be removed. Issue #33533
Removes the experimental Tags feature for Variables. Issue #33361
The InfoBox & FeatureInfoBox are now deprecated please use the Alert component instead with severity info. Issue #33352
getTimeZone function to be able to get the current selected timeZone. #31900, @mckn