Don't lock for all of Sync/stop/reload as that holds up /metrics and the
UI when they want a list of active/dropped targets. Instead take
advantage of the fact that Sync/stop/reload cannot be called
concurrently by the scrape Manager and lock just on the targets
themselves.
Signed-off-by: Brian Brazil <brian.brazil@robustperception.io>
* promtool: Calculate mint and maxt per test
Previously a single test that used a later eval time would make all
other tests in the file share the [mint, maxt] and potentially evaluate
far more samples than needed.
Fixes: #8019
Signed-off-by: David Leadbeater <dgl@dgl.cx>
* add networking.k8s.io for ingress
level=error ts=2020-10-19T08:32:30.544Z caller=klog.go:96 component=k8s_client_runtime func=ErrorDepth msg="github.com/prometheus/prometheus/discovery/kubernetes/kubernetes.go:494: Failed to watch *v1beta1.Ingress: failed to list *v1beta1.Ingress: ingresses.networking.k8s.io is forbidden: User \"system:serviceaccount:monitoring:prometheus\" cannot list resource \"ingresses\" in API group \"networking.k8s.io\" at the cluster scope"
Signed-off-by: root <likerj@inspur.com>
* Update rbac-setup.yml
Signed-off-by: root <likerj@inspur.com>
* dynamically determine path prefix
Signed-off-by: James Ranson <james_ranson@cable.comcast.com>
* minor changes per PR review
Signed-off-by: James Ranson <james_ranson@cable.comcast.com>
* use Context for apiPath and pathPrefix
Signed-off-by: James Ranson <james_ranson@cable.comcast.com>
* remove unhandled "/version" path
Signed-off-by: James Ranson <james_ranson@cable.comcast.com>
* only process index once instead of on every req
Signed-off-by: James Ranson <james_ranson@cable.comcast.com>
* remove unneeded tag fragment
Signed-off-by: James Ranson <james_ranson@cable.comcast.com>
* switch api path to const
Signed-off-by: James Ranson <james_ranson@cable.comcast.com>
* revert
Signed-off-by: James Ranson <james_ranson@cable.comcast.com>
* update tests
Signed-off-by: James Ranson <james_ranson@cable.comcast.com>
* linter updates
Signed-off-by: James Ranson <james_ranson@cable.comcast.com>
* simplify
Signed-off-by: James Ranson <james_ranson@cable.comcast.com>
* updates per peer review
Signed-off-by: James Ranson <james_ranson@cable.comcast.com>
When the latest tag is added, also add a tag for the major version
when the version tag matches the latest release defined in VERSION.
Signed-off-by: Victor Araujo <vear91@gmail.com>
* Close resources after failing to startup TSDB
Signed-off-by: arthursens <arthursens2005@gmail.com>
* Return close error instead of logging
Signed-off-by: arthursens <arthursens2005@gmail.com>
* Change named return's name
Signed-off-by: arthursens <arthursens2005@gmail.com>
This is how much memory we use to load in the on-disk
symbol tables, not the size of the tables themselves.
Signed-off-by: Brian Brazil <brian.brazil@robustperception.io>
This should be the way forward when importing libraries in jsonnet. It's
closer to how Go imports look and makes it more obvious where packages
live.
This is not breaking anything, as the old imports were already symlinks
to the now directly used directories.
Signed-off-by: Matthias Loibl <mail@matthiasloibl.com>
* Mixin: Ignore unset remote write timestamp
This pull request ignores the zero value of highest_sent_timestamp_seconds
in Highest Timestamp In vs. Highest Timestamp Sent which just show that
remote write has not been successful yet.
Signed-off-by: Julien Pivotto <roidelapluie@inuits.eu>
In #7399, an early validity check of the config was introduced to
prevent the scenario where an invalid config is only detected after a
possibly very long startup procedure. However, the respective success
metrics are not updated after the initial validation so that the
success metrics suggest an invalid config. If the startup procedure,
like replaying the WAL, really takes very long, alerts about invalid
config will trigger.
This commit sets the succes metrics after initial validation. They
will be set again after the "real" config (re-)load, but that
shouldn't be a problem. The metric now truthfully represents whenever
the config was successfully loaded, no matter if the result was then
thrown away (because it was just for validation) or actually used.
Signed-off-by: beorn7 <beorn@grafana.com>
This was already fixed by #8013, but add a test case anyway
in case the regexp engine changes in future.
Signed-off-by: Brian Brazil <brian.brazil@robustperception.io>
If an alert test had a failing test, then any other alert test interval
specified after that point would result in the test exiting early.
This made debugging some tests more difficult than needed.
Now only exit early for evaluation failures.
Signed-off-by: David Leadbeater <dgl@dgl.cx>
