* added backup locking/unlocking against retention policy to vmbackupmanager
Signed-off-by: Alexander Marshalov <_@marshalov.org>
* added docs for new commands
Signed-off-by: Alexander Marshalov <_@marshalov.org>
* fix review comments
Signed-off-by: Alexander Marshalov <_@marshalov.org>
---------
Signed-off-by: Alexander Marshalov <_@marshalov.org>
* feat: improvement of the top queries page
* vmui/docs: enhancements to top queries page
* Apply suggestions from code review
---------
Co-authored-by: Aliaksandr Valialkin <valyala@victoriametrics.com>
vmui: change default font size to 14px for better readability
vmui: fix bug with missing text on buttons in safari
---------
Co-authored-by: Roman Khavronenko <roman@victoriametrics.com>
* app/vmui: added Labels with the highest number of unique values
* app/vmui: cleanup
* app/vmui: cleanup
* app/vmui: add table description
* app/vmui: fix comment, updated CHANGELOG.md
* app/vmui: disable links
* app/vmui: added actions to the table, it will show values for selected label with the highest number of series
* app/vmui: fix comment
* vmalert: expand rule groups on anchor click
before, anchor click was only updating the URL.
To expand the group, user had to click on rule's block.
Now, group will toggle automatically.
* vmalert: allow filtering group in web UI
The new filter allows to filter groups and rules within
groups by: errors only or noMatch only.
The filtering supposed to help navigating big numbers of groups/rules.
Filtering is reflected in URL, so can be shared as a link.
Signed-off-by: hagen1778 <roman@victoriametrics.com>
Without reset, labels duplicates could have been added during stream aggregation.
Since `ctx.Labels` is reused during processing of many series, each series will
add its labels to the context. Even if the same labels were already addeded on prev
iteration. Now, we reset `ctx.Labels` on each iteration to contain so labels from
different series didn't interfere.
This could have cause exceeding of the limit on number of labels per pushed time series.
https://github.com/VictoriaMetrics/VictoriaMetrics/issues/4277
Signed-off-by: hagen1778 <roman@victoriametrics.com>
app/vmalert: detect alerting rules which don't match any series at all
vmalert starts to understand /query responses which contain object:
```
"stats":{"seriesFetched": "42"}
```
If object is present, vmalert parses it and populates a new field
`SeriesFetched`. This field is then used to populate the new metric
`vmalert_alerting_rules_last_evaluation_series_fetched` and to
display warnings in the vmalert's UI.
If response doesn't contain the new object (Prometheus or
VictoriaMetrics earlier than v1.90), then `SeriesFetched=nil`.
In this case, UI will contain no additional warnings.
And `vmalert_alerting_rules_last_evaluation_series_fetched` will
be set to `-1`. Negative value of the metric will help to compile
correct alerting rule in follow-up.
Thanks for the initial implementation to @Haleygo
See https://github.com/VictoriaMetrics/VictoriaMetrics/pull/4056
See https://github.com/VictoriaMetrics/VictoriaMetrics/issues/4039
Signed-off-by: hagen1778 <roman@victoriametrics.com>
It appears that 90% usage for anonymous mem usage
is already concerning. So we lowering the threshold to 80%.
Signed-off-by: hagen1778 <roman@victoriametrics.com>
When using `retentionTimezoneOffset` and having local timezone being more than 4 hours different from UTC indexdb retention calculation could return negative value. This caused indexdb rotation to get in loop.
Fix calculation of offset to use `retentionTimezoneOffset` value properly and add test to cover all legit timezone configs.
See:
- https://github.com/VictoriaMetrics/VictoriaMetrics/issues/4207
- https://github.com/VictoriaMetrics/VictoriaMetrics/pull/4206
Signed-off-by: Zakhar Bessarab <z.bessarab@victoriametrics.com>
Co-authored-by: Nikolay <nik@victoriametrics.com>
Windows doesn't allow to remove dir with opened files. Usually it's a case for snapshots, hard cannot be removed if file is openned.
With this change, dir will be renamed and properly deleted at the next process start.
It's recommended to restart vmstorage/vmsingle for snapshots deletion completion periodically.
https://github.com/VictoriaMetrics/VictoriaMetrics/issues/70
* lib/promscrape/discovery/kubernetes: add common labels to all ports discovered from endpoints
Sets
`__meta_kubernetes_endpoints_name` and `__meta_kubernetes_namespace` labels to all ports of pod.
Prometheus sets those labels to all ports in pod (0ab9553611/discovery/kubernetes/endpoints.go (L267C15-L269)) even if port is not matching any service.
See: #4154
Signed-off-by: Zakhar Bessarab <z.bessarab@victoriametrics.com>
* lib/promscrape/discovery/kubernetes: fix test for updated discovery logic
Signed-off-by: Zakhar Bessarab <z.bessarab@victoriametrics.com>
---------
Signed-off-by: Zakhar Bessarab <z.bessarab@victoriametrics.com>
Properly return empty slices instead of nil for `/api/v1/rules` and `/api/v1/alerts` API handlers.
This improves compatibility with Grafana.
https://github.com/VictoriaMetrics/VictoriaMetrics/issues/4221
Signed-off-by: hagen1778 <roman@victoriametrics.com>
Supports using `**` for `-rule` and `-rule.templates`: `dir/**/*.tpl` loads contents of dir and all subdirectories recursively.
See: #4041
Signed-off-by: Zakhar Bessarab <z.bessarab@victoriametrics.com>
Co-authored-by: Artem Navoiev <tenmozes@gmail.com>
Co-authored-by: Nikolay <nik@victoriametrics.com>
* lib/promscrape: adds filter for consul_sd_configs:
it allows advanced filtering for consul service discovery requests
https://github.com/VictoriaMetrics/VictoriaMetrics/issues/4183
* typo fix
* removes deprecation mentions since it's not relevant
* Update docs/CHANGELOG.md
Co-authored-by: Roman Khavronenko <roman@victoriametrics.com>
---------
Co-authored-by: Roman Khavronenko <roman@victoriametrics.com>
Templating of `-external.alert.source` is not expected to have access to the query which was causing runtime error when query function was passed as nil.
See: #4181
Signed-off-by: Zakhar Bessarab <z.bessarab@victoriametrics.com>
* feat: display heatmap in the explore metrics (#4111)
* fix: correct calc step for heatmap
* fix: remove spaces in the result of getDurationFromMilliseconds
* feat: add button "show today" to date picker
* feat: add comparison with the prev day (#3967)
* vmui/docs: add comparison of data to cardinality page
* feat: add WithTemplate page
* app/vmselect/prometheus: enable json mode for expand with expr API
* app/vmselect/prometheus: enable CORS and add content type
* feat: add api for expand with templates
* fix: remove console from useExpandWithExprs
* app/vmselect/prometheus: fix escaping
* vmui: integrate WITH template
* app/vmctl: check content type instead of form param
* fix: add content-type for fetch with-exprs
* fix: add a header to the server's response that allows the "Content-Type" header
* app/vmctl: added comment and cleanup
* app/vmctl: use format query param
---------
Co-authored-by: dmitryk-dk <kozlovdmitriyy@gmail.com>
* app/vmctl: add support for the different time format in the native binary protocol
* app/vmctl: update flag description, update CHANGELOG.md
* app/vmctl: add comment to exported function
* lib/httpserver: introduce `-http.maxConcurrentRequests` command-line flag
Introduce `-http.maxConcurrentRequests` command-line flag to protect
VM components from resource exhaustion during unexpected spikes of HTTP requests.
By default, the new flag's value is set to 0 which means no limits are applied.
Signed-off-by: hagen1778 <roman@victoriametrics.com>
* lib/httpserver: mention http.maxConcurrentRequests in docs
Signed-off-by: hagen1778 <roman@victoriametrics.com>
---------
Signed-off-by: hagen1778 <roman@victoriametrics.com>
* vmalert: retry datasource requests with EOF or unexpected EOF errors
Retry failed read request on the closed connection one more time.
This may improve rules execution reliability when connection
between vmalert and datasource closes unexpectedly.
Signed-off-by: hagen1778 <roman@victoriametrics.com>
* vmalert: fix old tests
Signed-off-by: hagen1778 <roman@victoriametrics.com>
---------
Signed-off-by: hagen1778 <roman@victoriametrics.com>
This handler will instruct search engines that indexing is not allowed for the content exposed to the internet. This should help to address issues like #4128 when instances are exposed to the internet without authentication.
Improperly configured -bigMergeConcurrency command-line flag usually leads to uncontrolled
growth of unmerged parts, which, in turn, increases CPU usage and query durations.
So it is better deprecating this flag. In rare cases -smallMergeConcurrency command-line flag
can be used instead for controlling the concurrency of background merges.
This makes it easier to understand exact point in time which is included in this backup.
Signed-off-by: Zakhar Bessarab <z.bessarab@victoriametrics.com>
* lib/promscrape: fix the problem with scrape work duplicates when file_sd_config can't be read
* lib/promscrape: clarified comment
* lib/promscrape: made better approach to handle a problem with growing []*ScrapeWork on each error when loading config
* lib/promscrape: added CHANGELOG.md
* Update docs/CHANGELOG.md
---------
Co-authored-by: Aliaksandr Valialkin <valyala@victoriametrics.com>
* feat: add tips for working with the graph and legend
* feat: add the ability to collapse the legend
* vmui/docs: add the ability to collapse the legend
---------
Co-authored-by: Aliaksandr Valialkin <valyala@victoriametrics.com>
* lib/storage: check for free disk space before opening tables
We check for free disk space before call to `openTable`,
so `Storage` can be set to ReadOnly before mergeWorkers start.
Before the change, there was a chance that merges will start
even if Storage has to start in ReadOnly mode because of
`-storage.minFreeDiskSpaceBytes` limit.
https://github.com/VictoriaMetrics/VictoriaMetrics/issues/4023
Signed-off-by: hagen1778 <roman@victoriametrics.com>
* lib/storage: chore
Signed-off-by: hagen1778 <roman@victoriametrics.com>
* Update lib/storage/storage.go
---------
Signed-off-by: hagen1778 <roman@victoriametrics.com>
Co-authored-by: Aliaksandr Valialkin <valyala@victoriametrics.com>
- Make sure that the last successfully loaded config is used on hot-reload failure
- Properly cleanup resources occupied by already initialized aggregators
when the current aggregator fails to be initialized
- Expose distinct vmagent_streamaggr_config_reload* metrics per each -remoteWrite.streamAggr.config
This should simplify monitoring and debugging failed reloads
- Remove race condition at app/vminsert/common.MustStopStreamAggr when calling sa.MustStop() while sa
could be in use at realoadSaConfig()
- Remove lib/streamaggr.aggregator.hasState global variable, since it may negatively impact scalability
on system with big number of CPU cores at hasState.Store(true) call inside aggregator.Push().
- Remove fine-grained aggregator reload - reload all the aggregators on config change instead.
This simplifies the code a bit. The fine-grained aggregator reload may be returned back
if there will be demand from real users for it.
- Check -relabelConfig and -streamAggr.config files when single-node VictoriaMetrics runs with -dryRun flag
- Return back accidentally removed changelog for v1.87.4 at docs/CHANGELOG.md
Updates https://github.com/VictoriaMetrics/VictoriaMetrics/issues/3639
Verifying status code helps to avoid misleading errors caused by attempt to parse unsuccessful response.
Related issue: #4034
Signed-off-by: Zakhar Bessarab <z.bessarab@victoriametrics.com>
- Compare directory names instead of paths to directory when determining which persistent queues must be deleted
This is less error-prone solution, since paths to the same directory can differ, which could lead
to accidental directory removal for the existing -remoteWrite.url
- Log the `removed %d dangling queues` message when at least a single queue has been removed
- Consistently use filepath.Join() for creating paths to persistent queues.
This is needed for Windows support (see https://github.com/VictoriaMetrics/VictoriaMetrics/issues/70 )
- Clarify the description of the change at docs/CHANGELOG.md
Updates https://github.com/VictoriaMetrics/VictoriaMetrics/issues/4014
- Allocate and initialize seriesByWorkerID slice in a single go instead
of initializing every item in the list separately.
This should reduce CPU usage a bit.
- Properly set anti-false sharing padding at timeseriesWithPadding structure
- Document the change at docs/CHANGELOG.md
Updates https://github.com/VictoriaMetrics/VictoriaMetrics/issues/3966
* allowed using dashes and dots in environment variables names for templating config files with envtemplate (#3999)
Signed-off-by: Alexander Marshalov <_@marshalov.org>
* Apply suggestions from code review
---------
Signed-off-by: Alexander Marshalov <_@marshalov.org>
Co-authored-by: Aliaksandr Valialkin <valyala@victoriametrics.com>
* lib/netutil: log only parsing errors for proxy-protocol
Previosly every error was logged. With configured TCP health checks at load-balancer or kubernetes, vmauth spams a lot of false positive error message into logs
* Update docs/CHANGELOG.md
Co-authored-by: Roman Khavronenko <roman@victoriametrics.com>
* Update lib/netutil/tcplistener.go
Co-authored-by: Roman Khavronenko <roman@victoriametrics.com>
---------
Co-authored-by: Aliaksandr Valialkin <valyala@victoriametrics.com>
Co-authored-by: Roman Khavronenko <roman@victoriametrics.com>
This commit changes background merge algorithm, so it becomes compatible with Windows file semantics.
The previous algorithm for background merge:
1. Merge source parts into a destination part inside tmp directory.
2. Create a file in txn directory with instructions on how to atomically
swap source parts with the destination part.
3. Perform instructions from the file.
4. Delete the file with instructions.
This algorithm guarantees that either source parts or destination part
is visible in the partition after unclean shutdown at any step above,
since the remaining files with instructions is replayed on the next restart,
after that the remaining contents of the tmp directory is deleted.
Unfortunately this algorithm doesn't work under Windows because
it disallows removing and moving files, which are in use.
So the new algorithm for background merge has been implemented:
1. Merge source parts into a destination part inside the partition directory itself.
E.g. now the partition directory may contain both complete and incomplete parts.
2. Atomically update the parts.json file with the new list of parts after the merge,
e.g. remove the source parts from the list and add the destination part to the list
before storing it to parts.json file.
3. Remove the source parts from disk when they are no longer used.
This algorithm guarantees that either source parts or destination part
is visible in the partition after unclean shutdown at any step above,
since incomplete partitions from step 1 or old source parts from step 3 are removed
on the next startup by inspecting parts.json file.
This algorithm should work under Windows, since it doesn't remove or move files in use.
This algorithm has also the following benefits:
- It should work better for NFS.
- It fits object storage semantics.
The new algorithm changes data storage format, so it is impossible to downgrade
to the previous versions of VictoriaMetrics after upgrading to this algorithm.
Updates https://github.com/VictoriaMetrics/VictoriaMetrics/issues/3236
Updates https://github.com/VictoriaMetrics/VictoriaMetrics/issues/3821
Updates https://github.com/VictoriaMetrics/VictoriaMetrics/issues/70
* app/vmagent: allow vm proto for kafka consumer and producer
it should reduce network usage up to 50%.
According to benchmarks without any encoding at kafka topic, it reduces traffic up to 50%.
With enabled zstd at kafka topic, it shows no diffence in traffic. So it
doesn't make much sense to use it.
https://github.com/VictoriaMetrics/VictoriaMetrics/issues/1225
* mention eb61a7dd68b834b08d01727a918f207700348ada at changelog
* app/vmagent: bumps kafka lib version
it allows compiling vmagent for arm64 machines
fixes https://github.com/VictoriaMetrics/VictoriaMetrics/issues/2271
* mention d19b1a888248c96cfd7ccee00ba6f596d89be1d7 at change log
* app/vmagent: adds natural concurrency for kafka consumer
it should improve performance for data consumption
https://github.com/VictoriaMetrics/VictoriaMetrics/issues/1957
* mention change 0c143bb22ca2e7e0b7eec9bc84a94ee2b41626ca
* Update app/vmagent/kafka/consumer.go
Co-authored-by: Roman Khavronenko <roman@victoriametrics.com>
* Update app/vmagent/kafka/consumer_cgo.go
Co-authored-by: Roman Khavronenko <roman@victoriametrics.com>
---------
Co-authored-by: Aliaksandr Valialkin <valyala@victoriametrics.com>
Co-authored-by: Roman Khavronenko <roman@victoriametrics.com>
* vmalert: support concurrent reading from object storage
Config reading from GCS or S3 can be slow if object storage
contains a big number of files. Object storages are usually
fast for downloading and are slow for individual operations.
If there would be thousands of files to read, vmalert could
spend significant time for retrieving those because it is
done sequentially.
The change introduces ability to read configs from object
storage concurrently. By default, both GCS and S3 are now
read with 50 concurrent readers. This significantly reduces
the load time:
* loading 500 files with concurrency=1 takes 27s
* loading 500 files with concurrency=50 takes <1s
* vmalert: add note to Changelog
* vmalert: cleanup
* vmalert: use ticker properly
* app/vmalert: improve status reporting during config loading
* vmalert: support concurrent reading from object storage
Config reading from GCS or S3 can be slow if object storage
contains a big number of files. Object storages are usually
fast for downloading and are slow for individual operations.
If there would be thousands of files to read, vmalert could
spend significant time for retrieving those because it is
done sequentially.
The change introduces ability to read configs from object
storage concurrently. By default, both GCS and S3 are now
read with 50 concurrent readers. This significantly reduces
the load time:
* loading 500 files with concurrency=1 takes 27s
* loading 500 files with concurrency=50 takes <1s
* app/vmalert: make linter happy
The change also introduces `List` method to `FS` interface.
The `List` method can be used for wildcard support in object storage FS.
Signed-off-by: hagen1778 <roman@victoriametrics.com>
Co-authored-by: Nikolay <nik@victoriametrics.com>
- Sync the description for -httpListenAddr.useProxyProtocol command-line flag at vmagent and vmauth,
so it is consistent with the description at vmauth and victoria-metrics
- Add a sample of panic text to docs/CHANGELOG.md, so it could be googled
- Mention the -httpListenAddr.useProxyProtocol command-line flag in the description for the bugfix
Updates https://github.com/VictoriaMetrics/VictoriaMetrics/issues/3335
app/vmctl: vm-native - split migration on per-metric basis
`vm-native` mode now splits the migration process on per-metric basis.
This allows to migrate metrics one-by-one according to the specified filter.
This change allows to retry export/import requests for a specific metric and provides a better
understanding of the migration progress.
---------
Signed-off-by: hagen1778 <roman@victoriametrics.com>
Co-authored-by: hagen1778 <roman@victoriametrics.com>
lib{mergset,storage}: prevent possible race condition with logging stats for merges
Previously partwrapper could be release by background process and reference for part may be invalid
during logging stats. It will lead to panic at vmstorage
https://github.com/VictoriaMetrics/VictoriaMetrics/issues/3897
- Use flag.Duration instead of flagutil.Duration for -snapshotCreateTimeout,
since the flagutil.Duration is intended mostly for big durations, e.g. days, months and years,
while the -snapshotCreateTimeout is usually smaller than one hour.
- Add links to https://docs.victoriametrics.com/#how-to-work-with-snapshots in docs/CHANGELOG.md,
so readers could easily find the corresponding docs when reading the changelog.
- Properly remove all the created directories on unsuccessful attempt to create
snapshot in Storage.CreateSnapshot().
Updates https://github.com/VictoriaMetrics/VictoriaMetrics/issues/3551
* lib/{fs,mergeset,storage}: skip `.must-remove.` dirs when creating snapshot (#3858)
* lib/{mergeset,storage}: add timeout configuration for snapshots creation, remove incomplete snapshots from storage
* docs: fix formatting
* app/vmstorage: add metrics to track status of snapshots
* app/vmstorage: use `vm_http_requests_total` metric for snapshot endpoints metrics, rename new flag to make name more clear
Signed-off-by: Zakhar Bessarab <z.bessarab@victoriametrics.com>
* app/vmstorage: update flag name in docs
Signed-off-by: Zakhar Bessarab <z.bessarab@victoriametrics.com>
* app/vmstorage: reflect new metrics names change in docs
Signed-off-by: Zakhar Bessarab <z.bessarab@victoriametrics.com>
---------
Signed-off-by: Zakhar Bessarab <z.bessarab@victoriametrics.com>
Co-authored-by: Aliaksandr Valialkin <valyala@victoriametrics.com>
* lib/promscrape: set `vm_promscrape_config_last_reload_successful` to 1 if there was no promscrape config provided
Signed-off-by: Zakhar Bessarab <z.bessarab@victoriametrics.com>
* lib/promscrape: register `vm_promscrape_config_*` metrics only in case promscrape config is used
Signed-off-by: Zakhar Bessarab <z.bessarab@victoriametrics.com>
---------
Signed-off-by: Zakhar Bessarab <z.bessarab@victoriametrics.com>
Co-authored-by: Aliaksandr Valialkin <valyala@victoriametrics.com>
* vmselect/promql: check for deadline in `count_values` fn
`count_values` could be very slow during the data processing.
Checking for deadline between iterations supposed to reduce
probability of exceeding `search.maxQueryDuration`.
The change also adds a new trace record, which captures the time
spent in aggregation function. Before that, the trace for aggr funcs
could be confusing since it doesn't account for all the places where
time was spent.
Signed-off-by: hagen1778 <roman@victoriametrics.com>
* wip
---------
Signed-off-by: hagen1778 <roman@victoriametrics.com>
Co-authored-by: Aliaksandr Valialkin <valyala@victoriametrics.com>
* metricsql: support optional 2nd argument for rollup functions
Support optional 2nd argument `min`, `max` or `avg` for rollup functions:
* rollup
* rollup_delta
* rollup_deriv
* rollup_increase
* rollup_rate
* rollup_scrape_interval
If second argument is passed, then rollup function will return only the selected aggregation type.
This change can be useful for situations where only one type of rollup calculation is needed.
For example, `rollup_rate(requests_total[5m], "max")`.
Signed-off-by: hagen1778 <roman@victoriametrics.com>
* wip
---------
Signed-off-by: hagen1778 <roman@victoriametrics.com>
Co-authored-by: Aliaksandr Valialkin <valyala@victoriametrics.com>
- Return immediately on context cancel during the backoff sleep.
This should help with https://github.com/VictoriaMetrics/VictoriaMetrics/issues/3747
- Add a comment describing why the second attempt to obtain the response from remote side
is perfromed immediately after the first attempt.
- Remove fasthttp dependency from lib/promscrape/discoveryutils
- Set context deadline before calling doRequestWithPossibleRetry().
This simplifies the doRequestWithPossibleRetry() a bit.
Updates https://github.com/VictoriaMetrics/VictoriaMetrics/issues/3293
* fix: do not use exponential backoff for first retry of scrape request (#3293)
* lib/promscrape: refactor `doRequestWithPossibleRetry` backoff to simplify logic
Signed-off-by: Zakhar Bessarab <z.bessarab@victoriametrics.com>
* Update lib/promscrape/client.go
Co-authored-by: Roman Khavronenko <roman@victoriametrics.com>
* lib/promscrape: refactor `doRequestWithPossibleRetry` to make it more straightforward
Signed-off-by: Zakhar Bessarab <z.bessarab@victoriametrics.com>
---------
Signed-off-by: Zakhar Bessarab <z.bessarab@victoriametrics.com>
Co-authored-by: Roman Khavronenko <roman@victoriametrics.com>
- Do not generate __meta_server label, since it is unavailable in Prometheus.
- Add a link to https://docs.victoriametrics.com/sd_configs.html#kuma_sd_configs to docs/CHANGELOG.md,
so users could click it and read the docs without the need to search the corresponding docs.
- Remove kumaTarget struct, since it is easier generating labels for discovered targets
directly from the response returned by Kuma. This simplifies the code.
- Store the generated labels for discovered targets inside atomic.Value. This allows reading them
from concurrent goroutines without the need to use mutex.
- Use synchronouse requests to Kuma instead of long polling, since there is a little sense
in the long polling when the Kuma server may return 304 Not Modified response every -promscrape.kumaSDCheckInterval.
- Remove -promscrape.kuma.waitTime command-line flag, since it is no longer needed when long polling isn't used.
- Set default value for -promscrape.kumaSDCheckInterval to 30s in order to be consistent with Prometheus.
- Remove unnecessary indirections for string literals, which are used only once, in order to improve code readability.
- Remove unused fields from discoveryRequest and discoveryResponse.
- Update tests.
- Document why fetch_timeout and refresh_interval options are missing in kuma_sd_config.
- Add docs to discoveryutils.RequestCallback and discoveryutils.ResponseCallback,
since these are public types.
Side notes: it is weird that Prometheus implementation for kuma_sd_configs sets `instance` label,
since usually this label is set by the Prometheus itself to __address__ after the relabeling phase.
See https://www.robustperception.io/life-of-a-label/
Updates https://github.com/VictoriaMetrics/VictoriaMetrics/issues/3389
See https://github.com/prometheus/prometheus/issues/7919
and https://github.com/prometheus/prometheus/pull/8844
as a reference implementation in Prometheus
