6.2-beta1 changed so that data sources, dashboards and alert notifiers
are provisioned on service Init where's before only data sources and
alert notifiers was provisioned on service Init and dashboards was first
provisioned in service Run which happens after Grafana server have
started.
This change reverts so that dashboard provisioning are first happening
in service Run, after Grafana server have started.
Fixes#21133
* svc alerting - use a shorter ctx to upload the img
This will prevent timeout on img upload to cancel the notifications from being sent
* components img uploader - pass the ctx to aws lib
* make webdavuploader use the ctx
* make azureblobuploader use the ctx
* rename uploadImage() to renderAndUploadImage()
for better clarity about what this method work
* Use timeout + 2s for plugin renderer (same as service and phantomjs)
* Make sure that original EvalContext is updated after render and upload
* Verify notification sent even if render or image upload times out
* fix lint
* fixes after review
Co-authored-by: Edouard Hur <3418467+hekmon@users.noreply.github.com>
Fixes#21018
Adds support for collecting metrics from backend plugins and
exposing them thru Grafana's Prometheus metrics endpoint.
Enables to check health of backend plugin by using the route
`/api/plugins/<plugin id>/health`.
Uses sdk v0.6.0.
Closes#20984
#21350 introduced a bug regarding import of plugin dashboards.
This should fix this and add custom validation if not importing
plugin dashboard and dashboard property is missing.
Ref #21350
Co-Authored-By: Arve Knudsen <arve.knudsen@gmail.com>
Breaking change: If disabled the cookie samesite cookie attribute
will not be set, but if none the attribute will be set and is a
breaking change compared to before where none did not render the
attribute. This was due to a known issue in Safari.
Co-Authored-By: Arve Knudsen <arve.knudsen@gmail.com>
Co-Authored-By: Diana Payton <52059945+oddlittlebird@users.noreply.github.com>
Fixes#19847
When an alert is sent by e-mail, the process sends an e-mail to
each recipient separately. This PR is a single delivery to all recipients.
For companies that use e-mail extensively, this is necessary in order
not to overload the sending queue.
Replaces #18013Fixes#12650
Co-authored-by: Henrique Oliveira <holiiveira@users.noreply.github.com>
Moving info log from phantomjs to rendering service
so it's logged for all kinds of renderers.
Add debug log for image renderer plugin and remote
renderer.
Adds support for using custom dockerfiles to pre-install image
renderer plugins.
Updates docs for docker and image rendering.
Fixes#20241
Co-Authored-By: Leonard Gram <leo@xlson.com>
Backend plugins is recommended to use hclog with json
formatting to get proper log output in grafana server log.
Old hclog-wrapper.go that I deleted while back is still in
the repo so deletes that.
When using Raw Document query with Elasticsearch there's a special
response from datasource that is used which includes a type field with
the value json. In the table panel there is a transformation for JSON
data which up until this fix didn't work at all due to the new data
structure we call data frames.
Co-Authored-By: Hugo Häggmark <hugo.haggmark@grafana.com>
Fixes#19531
* use grafana-plugin-model dependency that uses go modules
* use grafana-plugin-model with updated hashicorp/go-plugin
* use grafana-plugin-model with re-compiled protos
* test using protoc-gen-go v1.2.0 tag
* use grafana-plugin-model with re-compiled protos
* chore: fix deprecation warning for lint
* use latest grafana-plugin-model
Fixes#19454
Only return certain mysql errors from backend.
The following errors is returned as is from backend:
error code 1064 (parse error)
error code 1054 (bad column/field selected)
error code 1146 (table not exists)
Any other errors is logged and returned as a generic
error.
Restrict use of certain functions:
Do not allow usage of the following in query:
system_user()
session_user()
current_user() or current_user
user()
show grants
Fixes#19360
* build: make sign rpm packages not depend on checking out private key
* build: move commands from circleci config into verify signed packages script
* build: split update and publish of deb and rpm into two scripts
* use files argument for sign and verify packages
* validate files argument for sign and verify packages
* update test publish of deb/rpm readme
This reverts commit 2514209 from #18384. Reason is that it doesn't
work due to xorm 0.7.1 which doesn't support this new connectionstring
format.
Fixes#19189
Ref #18384
Ref #17665
Adds a new "Image Rendering" page in Administration section.
Updates configuration page with rendering settings and also
default.ini and sample.ini.
Updates and cleanup pages that referencing image rendering.
Ref #18914
Don't update total stats metrics if reporting is disabled.
New setting disable_total_stats for turning off update
of total stats (stat_totals_*) metrics.
Ref #19137
Adds a new requirements page to the installation docs.
Wanted/needed some place to document known browser
issues and thought that adding this missing page together
with OS, hardware, database and browser requirements
would be a good thing.
Fix: move 6.4 section in upgrade notes further down.
Fixes#16487
Ref #18690
Existing /api/alert-notifications now requires at least editor access.
Existing /api/alert-notifiers now requires at least editor access.
New /api/alert-notifications/lookup returns less information than
/api/alert-notifications and can be access by any authenticated user.
Existing /api/org/users now requires org admin role.
New /api/org/users/lookup returns less information than
/api/org/users and can be access by users that are org admins,
admin in any folder or admin of any team.
UserPicker component now uses /api/org/users/lookup instead
of /api/org/users.
Fixes#17318
Elasticsearch v7.0 changed the behavior of max concurrent shard
requests and the default to 5. v5.6 and before 7.0 the default
is 256. This adds some additional behavior given certain
version is selected when configure the datasource to set
default max concurrent shard requests.
Changing from a version pre-v7.0+ to v7.0+ sets max
concurrent shard requests to 5.
Changing from a version v7.0+ to a pre-v7.0 sets max
concurrent shard requests to 256.
Fixes#17454
Adds version note for alert rule tags.
Adds version note for graph data links.
Adds information about secret_key and encryption format in use
Fixes#17815Fixes#17803
* tsdb: add support for setting debug flag of tsdb query
* alerting: adds debug flag in eval context
Debug flag is set when testing an alert rule and this debug
flag is used to return more debug information in test aler rule
response. This debug flag is also provided to tsdb queries so
datasources can optionally add support for returning additional
debug data
* alerting: improve test alert rule ui
Adds buttons for expand/collapse json and copy json to clipboard,
very similar to how the query inspector works.
* elasticsearch: implement support for tsdb query debug flag
* elasticsearch: embedding client response in struct
* alerting: return proper query model when testing rule
Adds a first version of project documentation/guideline for triaging
of issues. This adds some additional content and context so that
the issue triage process can be understood by both maintainers
and existing/new contributors and other interested parties.
Co-Authored-By: Andrej Ocenas <mr.ocenas@gmail.com>
Closes#17648
xorm introduced some changes in
https://github.com/go-xorm/xorm/pull/824 and
https://github.com/go-xorm/xorm/pull/876 which by default will use
public as the postgres schema and this was a breaking change compared
to before. Grafana has implemented a custom postgres dialect so above
changes wasn't a problem here. However, Grafana's custom database
migration was using xorm dialect to check if the migration table exists
or not.
For those using a custom search_path (schema) in postgres configured on
server, database or user level the migration table check would not find
the migration table since it was looking in public schema due to xorm
changes above. This had the consequence that Grafana's database
migration failed the second time since migration had already run
migrations in another schema.
This change will make xorm use an empty default schema for postgres and
by that mimic the functionality of how it was functioning before
xorm's changes above.
Fixes#16720
Co-Authored-By: Carl Bergquist <carl@grafana.com>
If a field or a label named level is returned from datasource that is
used as log level for the logs result instead of parsing the log level
from the message.
Closes#17122
Marcus Efraimsson
