opendev/grafyaml
Code Issues Proposed changes

Use fixed datasource UID

Browse Source 
As described inline, datasources now have a UID.  Set this to a fixed
hash of the URL to make dashboards which refer to datasources
explicitly portable.

Change-Id: I53e2aec7f635e8ce8793abb5755eccd2e6b3e4c5
This commit is contained in:
Ian Wienand 2022-06-29 14:25:08 +10:00
parent 8c973ef54b
commit 1db873f6f4
1 changed files with 39 additions and 0 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

39
grafana_dashboards/grafana/datasource.py
View File

@ -12,12 +12,16 @@
# License for the specific language governing permissions and limitations
# under the License.
import logging
import hashlib
import json
from requests import exceptions
from grafana_dashboards.grafana import utils
LOG = logging.getLogger(__name__)
class Datasource(object):
@ -39,6 +43,41 @@ class Datasource(object):
if self.is_datasource(name):
raise Exception('datasource[%s] already exists' % name)
# Always create this datasource with an fixed UID (*not* id)
# that is consistent across installations by hashing the
# name/url. A story why:
#
# Since ~ Grafana 8.3 or so datasources got a UID and all
# metrics now refer to their datasource by UID.
#
# Grafana can make "exportable" dashboards that replace the
# datasource UID with a variable (read about that at [1]).
# This is what we want ... the only problem is you can not
# automatically import such a dashboard -- it is a UI driven
# process where when clicking on "import" you end up in a
# wizard where you select the datasource and behind the scenes
# it goes and fills in the variables for you before saving the
# dashboard. You can search around the forums and github
# issues for where people are discussing this; the short story
# is that this happens in the UI via an undocumented
# api/dashboards/import call which upstream so far (June 2022)
# have no plans to export [2].
#
# So, by fixing the UID here, we ensure that dashboards are
# portable within the "grafyaml" ecosystem. i.e. if you used
# grafyaml to setup the datasource, be that in testing,
# locally or in production, you'll have a consistent
# datasource UID and your dashboards will work when imported
# to any other.
#
# [1] https://grafana.com/docs/grafana/latest/dashboards/export-import/
# [2] https://github.com/grafana/grafana/ \
# issues/9812#issuecomment-343216975
data['uid'] = hashlib.sha256(
data['url'].encode('utf-8')).hexdigest()[0:10]
LOG.debug('Setting UID of datasource %s to %s' %
(data['url'], data['uid']))
res = self.session.post(
self.url, data=json.dumps(data))