When Prometheus scrapes a target and it sees the same time series repeated multiple times it currently silently ignores that. This change adds a test for that and fixes the scrape loop so that:
* Only first sample for each unique time series is appended
* Duplicated samples increment the prometheus_target_scrapes_sample_duplicate_timestamp_total metric
This allows one to identify such scrape jobs and targets.
Also fix some tests and benchmark.
* add custom buckets to native histogram model
* simple copy for custom bounds
* return errors for unsupported add/sub operations
* add test cases for string and update appendhistogram in scrape to account for new schema
* check fields which are supposed to be unused but may affect results in equals
* allow appending custom buckets histograms regardless of max schema
Signed-off-by: Jeanette Tan <jeanette.tan@grafana.com>
This call bloats checkAddError signature and logic, we can and should call it from the main scrape logic.
Signed-off-by: Łukasz Mierzwa <l.mierzwa@gmail.com>
This test calls Rollback() which is normally called from within append code.
Doing so means that staleness tracking data is outdated and need to by cycled manually.
Signed-off-by: Łukasz Mierzwa <l.mierzwa@gmail.com>
OpenMetrics requires EOF comment at the end of metrics body, but the makeTestMetrics() function doesn't append it.
This means this benchmark tests a response with errors but I don't think that was the intention.
Signed-off-by: Łukasz Mierzwa <l.mierzwa@gmail.com>
When the scrape tolerance is bigger than 1% of the scrape interval, take
1% of the scrape interval as the tolerance instead of not aligning the
scrape at all.
Signed-off-by: Julien Pivotto <roidelapluie@o11y.eu>
This adds support for the new grammar of `{"metric_name", "l1"="val"}` to promql and some of the exposition formats.
This grammar will also be valid for non-UTF-8 names.
UTF-8 names will not be considered valid unless model.NameValidationScheme is changed.
This does not update the go expfmt parser in text_parse.go, which will be addressed by https://github.com/prometheus/common/issues/554/.
Part of https://github.com/prometheus/prometheus/issues/13095
Signed-off-by: Owen Williams <owen.williams@grafana.com>
Normally, a NaN value is never equal to any other value. Compare sample
values via `Float64bits` so that NaN values which are exactly the same
will compare equal.
Signed-off-by: Bryan Boreham <bjboreham@gmail.com>
Native Histograms: support native_histogram_min_bucket_factor in scrape_config
---------
Signed-off-by: Ziqi Zhao <zhaoziqi9146@gmail.com>
Signed-off-by: Björn Rabenstein <github@rabenste.in>
Co-authored-by: George Krajcsovits <krajorama@users.noreply.github.com>
Co-authored-by: Björn Rabenstein <github@rabenste.in>
They are used in multiple repos, so common is a better place for them.
Several packages now don't depend on `model/textparse`, e.g.
`storage/remote`.
Also remove `metadata` struct from `api.go`, since it was identical to
a struct in the `metadata` package.
Signed-off-by: Bryan Boreham <bjboreham@gmail.com>
* Append created timestamps.
Signed-off-by: Arthur Silva Sens <arthur.sens@coralogix.com>
* Log when created timestamps are ignored
Signed-off-by: Arthur Silva Sens <arthur.sens@coralogix.com>
* Proposed changes to Append CT PR.
Changes:
* Changed textparse Parser interface for consistency and robustness.
* Changed CT interface to be more explicit and handle validation.
* Simplified test, change scrapeManager to allow testability.
* Added TODOs.
Signed-off-by: bwplotka <bwplotka@gmail.com>
* Updates.
Signed-off-by: bwplotka <bwplotka@gmail.com>
* Addressed comments.
Signed-off-by: bwplotka <bwplotka@gmail.com>
* Refactor head_appender test
Signed-off-by: Arthur Silva Sens <arthur.sens@coralogix.com>
* Fix linter issues
Signed-off-by: Arthur Silva Sens <arthur.sens@coralogix.com>
* Use model.Sample in head appender test
Signed-off-by: Arthur Silva Sens <arthur.sens@coralogix.com>
---------
Signed-off-by: Arthur Silva Sens <arthur.sens@coralogix.com>
Signed-off-by: bwplotka <bwplotka@gmail.com>
Co-authored-by: bwplotka <bwplotka@gmail.com>
The 'ToFloat' method on integer histograms currently allocates new memory
each time it is called.
This commit adds an optional *FloatHistogram parameter that can be used
to reuse span and bucket slices. It is up to the caller to make sure the
input float histogram is not used anymore after the call.
Signed-off-by: Filip Petkovski <filip.petkovsky@gmail.com>
The scrape loop and scrape cache should use the same instance.
This brings the tests' behavior more in line with production.
Signed-off-by: Paulin Todev <paulin.todev@gmail.com>
Previously we had one per scrapePool, and one of those per configured
scraping job. Each pool holds a few unused buffers, so sharing one
across all scrapePools reduces total heap memory.
Signed-off-by: Bryan Boreham <bjboreham@gmail.com>
Too confusing to have `MetadataList` and `ListMetadata`, etc.
I standardised on the ones which are in an interface.
Signed-off-by: Bryan Boreham <bjboreham@gmail.com>
Since everything was serialized on a single mutex, it's exactly the same
if we process targets in sequence without starting goroutines.
Signed-off-by: Bryan Boreham <bjboreham@gmail.com>
Currently Prometheus will always request gzip compression from the target when sending scrape requests.
HTTP compression does reduce the amount of bytes sent over the wire and so is often desirable.
The downside of compression is that it requires extra resources - cpu & memory.
This also affects the resource usage on the target since it has to compress the response
before sending it to Prometheus.
This change adds a new option to the scrape job configuration block: enable_compression.
The default is true so it remains the same as current Prometheus behaviour.
Setting this option to false allows users to disable compression between Prometheus
and the scraped target, which will require more bandwidth but it lowers the resource
usage of both Prometheus and the target.
Fixes#12319.
Signed-off-by: Łukasz Mierzwa <l.mierzwa@gmail.com>
Fix and improve ingesting exemplars for native histograms.
See code comment for a detailed explanation of the algorithm.
Note that this changes the current behavior for all kind of samples slightly: We now allow exemplars with the same timestamp as during the last scrape if the value or the labels have changed.
Also note that we now do not ingest exemplars without timestamps for native histograms anymore.
Signed-off-by: Jeanette Tan <jeanette.tan@grafana.com>
Signed-off-by: György Krajcsovits <gyorgy.krajcsovits@grafana.com>
Co-authored-by: Björn Rabenstein <github@rabenste.in>
---------
Signed-off-by: Jeanette Tan <jeanette.tan@grafana.com>
Signed-off-by: György Krajcsovits <gyorgy.krajcsovits@grafana.com>
Signed-off-by: zenador <zenador@users.noreply.github.com>
Co-authored-by: György Krajcsovits <gyorgy.krajcsovits@grafana.com>
Co-authored-by: Björn Rabenstein <github@rabenste.in>
This commit introduces an additional test in `scrape_test.go` to verify
staleness tracking when `trackTimestampStaleness` is enabled. The new
`TestScrapeLoopAppendStalenessIfTrackTimestampStaleness` function
asserts that the scrape loop correctly appends staleness markers when
necessary, reflecting the expected behavior with the feature flag turned
on.
The previous tests were only testing end of scrape staleness.
Signed-off-by: Julien Pivotto <roidelapluie@o11y.eu>
* Document le and quantile label transition due to native histograms
Fixes: #12984
For full explanation see the related issue. The le and quantile labels
are formatted as float with trailing .0 for whole number values when
native histograms is enabled, e.g. 10.0. This changes the resulting series
in Prometheus if previously we scraped the whole number itself, e.g. 10
over the text format.
Signed-off-by: György Krajcsovits <gyorgy.krajcsovits@grafana.com>
---------
Signed-off-by: György Krajcsovits <gyorgy.krajcsovits@grafana.com>
Signed-off-by: George Krajcsovits <krajorama@users.noreply.github.com>
* A registerer is passed to the scrape Manager,
and all scrape metrics register with it.
* For now the registry which we pass to the scrape
Manager is still the global one.
Signed-off-by: Paulin Todev <paulin.todev@gmail.com>
* Added ability to specify scrape protocols to accept during HTTP content type negotiation.
This is done via new option in GlobalConfig and ScrapeConfig: "scrape_protocol"
Signed-off-by: bwplotka <bwplotka@gmail.com>
* Fixed readability and log message.
Signed-off-by: bwplotka <bwplotka@gmail.com>
---------
Signed-off-by: bwplotka <bwplotka@gmail.com>
We don't need the buffer to read the response until the scrape http call
returns; creating it earlier makes the buffer pool larger.
I split `scrape()` into `scrape()` which returns with the http response,
and `readResponse()` which decompresses and copies the data into the
supplied buffer. This design was chosen to minimize impact on the logic.
Signed-off-by: Bryan Boreham <bjboreham@gmail.com>
`lset` escapes to heap due to being passed through the text-parser
interface, so we can reduce garbage by hoisting it out of the loop so
only one allocation is done for every series in a scrape.
Signed-off-by: Bryan Boreham <bjboreham@gmail.com>
The parsing doesn't seem to be perfect as I don't get all classic buckets
possibly another bug found?
Signed-off-by: György Krajcsovits <gyorgy.krajcsovits@grafana.com>
It's possible (quite common on Kubernetes) to have a service discovery
return thousands of targets then drop most of them in relabel rules.
The main place this data is used is to display in the web UI, where
you don't want thousands of lines of display.
The new limit is `keep_dropped_targets`, which defaults to 0
for backwards-compatibility.
Signed-off-by: Bryan Boreham <bjboreham@gmail.com>
Continue to remove confusion that histogram samples are also samples
and histogram values are also values etc. by renaming float values and
float samples using the same schema as for histograms.
Concretely:
- result → resultFloats (corresponding to resultHistograms)
- pendingResult → pendingFloats (corresponding to pendingHistograms)
- rolledbackResult → rolledbackFloats (corresponding to rolledbackHistograms)
- sample → floatSample (corresponding to histogramSample)
This also order the fields in `collectResultAppender` more
consistently.
Signed-off-by: beorn7 <beorn@grafana.com>
This has become a requirement for native histograms, as a single
histogram sample commonly has many buckets, so that providing many
exemplars makes sense.
Since OM text doesn't support native histograms yet, the test had to
be expanded to also support protobuf test cases.
Signed-off-by: beorn7 <beorn@grafana.com>
I had changed the naming from "jitter" to "offset" in:
cb045c0e4b
...but I forgot to add this file to the commit to complete the renaming,
doing that now.
Signed-off-by: Julius Volz <julius.volz@gmail.com>
In digital communication, "jitter" usually refers to how much a signal deviates
from true periodicity, see https://en.wikipedia.org/wiki/Jitter. The way we are
using the "jitterSeed" in Prometheus does not affect the true periodicity at
all, but just introduces a constant phase shift (or offset) within the period.
So it would be more correct and less confusing to call the "jitterSeed" an
"offsetSeed" instead.
Signed-off-by: Julius Volz <julius.volz@gmail.com>
So far, if a target exposes a histogram with both classic and native
buckets, a native-histogram enabled Prometheus would ignore the
classic buckets. With the new scrape config option
`scrape_classic_histograms` set, both buckets will be ingested,
creating all the series of a classic histogram in parallel to the
native histogram series. For example, a histogram `foo` would create a
native histogram series `foo` and classic series called `foo_sum`,
`foo_count`, and `foo_bucket`.
This feature can be used in a migration strategy from classic to
native histograms, where it is desired to have a transition period
during which both native and classic histograms are present.
Note that two bugs in classic histogram parsing were found and fixed
as a byproduct of testing the new feature:
1. Series created from classic _gauge_ histograms didn't get the
_sum/_count/_bucket prefix set.
2. Values of classic _float_ histograms weren't parsed properly.
Signed-off-by: beorn7 <beorn@grafana.com>
Consider code like:
for i := 0; i < numTargets; i++ {
stopFuncs = append(stopFuncs, func() {
time.Sleep(i*20*time.Millisecond)
})
}
Because the loop variable i is shared by all closures,
all the stopFuncs sleep for numTargets*20 ms.
If the i were made per-iteration, as we are considering
for a future Go release, the stopFuncs would have sleep
durations ranging from 0 to (numTargets-1)*20 ms.
Two tests had code like this and were checking that the
aggregate sleep was at least numTargets*20 ms
("at least as long as the last target slept"). This is only true
today because i == numTarget during all the sleeps.
To keep the code working even if the semantics of this loop
change, this PR computes
d := time.Duration((i+1)*20) * time.Millisecond
outside the closure (but inside the loop body), and then each
closure has its own d. Now the sleeps range from 20 ms
to numTargets*20 ms, keeping the test passing
(and probably behaving closer to the intent of the test author).
The failure being fixed can be reproduced by using the current
Go development branch with
GOEXPERIMENT=loopvar go test
Signed-off-by: Russ Cox <rsc@golang.org>
Refactors textparser test to use a common test utility to create
protobuf representation from MetricFamily
Signed-off-by: György Krajcsovits <gyorgy.krajcsovits@grafana.com>
Wiser coders than myself have come to the conclusion that a `switch`
statement is almost always superior to a statement that includes any
`else if`.
The exceptions that I have found in our codebase are just these two:
* The `if else` is followed by an additional statement before the next
condition (separated by a `;`).
* The whole thing is within a `for` loop and `break` statements are
used. In this case, using `switch` would require tagging the `for`
loop, which probably tips the balance.
Why are `switch` statements more readable?
For one, fewer curly braces. But more importantly, the conditions all
have the same alignment, so the whole thing follows the natural flow
of going down a list of conditions. With `else if`, in contrast, all
conditions but the first are "hidden" behind `} else if `, harder to
spot and (for no good reason) presented differently from the first
condition.
I'm sure the aforemention wise coders can list even more reasons.
In any case, I like it so much that I have found myself recommending
it in code reviews. I would like to make it a habit in our code base,
without making it a hard requirement that we would test on the CI. But
for that, there has to be a role model, so this commit eliminates all
`if else` occurrences, unless it is autogenerated code or fits one of
the exceptions above.
Signed-off-by: beorn7 <beorn@grafana.com>
We haven't updated golint-ci in our CI yet, but this commit prepares
for that.
There are a lot of new warnings, and it is mostly because the "revive"
linter got updated. I agree with most of the new warnings, mostly
around not naming unused function parameters (although it is justified
in some cases for documentation purposes – while things like mocks are
a good example where not naming the parameter is clearer).
I'm pretty upset about the "empty block" warning to include `for`
loops. It's such a common pattern to do something in the head of the
`for` loop and then have an empty block. There is still an open issue
about this: https://github.com/mgechev/revive/issues/810 I have
disabled "revive" altogether in files where empty blocks are used
excessively, and I have made the effort to add individual
`// nolint:revive` where empty blocks are used just once or twice.
It's borderline noisy, though, but let's go with it for now.
I should mention that none of the "empty block" warnings for `for`
loop bodies were legitimate.
Signed-off-by: beorn7 <beorn@grafana.com>
It took a `Labels` where the memory could be re-used, but in practice
this hardly ever benefitted. Especially after converting `relabel.Process`
to `relabel.ProcessBuilder`.
Comparing the parameter to `nil` was a bug; `EmptyLabels` is not `nil`
so the slice was reallocated multiple times by `append`.
Lastly `Builder.Labels()` now estimates that the final size will depend
on labels added and deleted.
Signed-off-by: Bryan Boreham <bjboreham@gmail.com>
Since the Target object was just created in this function, nobody else
has a reference to it and there are no concerns about it being modified
concurrently so we don't need to copy the value.
Signed-off-by: Bryan Boreham <bjboreham@gmail.com>
Includes a rewrite of `resolveConflictingExposedLabels` to use
`labels.Builder.Get`, which simplifies it considerably.
Signed-off-by: Bryan Boreham <bjboreham@gmail.com>
Scraping targets are synced by creating the full set, then adding/removing any which have changed.
This PR speeds up the process of creating the full set.
I added a benchmark for `TargetsFromGroup`; it uses configuration from a typical Kubernetes SD.
The crux of the change is to do relabeling inside labels.Builder instead of converting to labels.Labels and back again for every rule. The change is broken into several commits for easier review.
This is a breaking change to `scrape.PopulateLabels()`, but `relabel.Process` is left as-is, with a new `relabel.ProcessBuilder` option.
Common service discovery mechanisms such as Kubernetes can generate a
lot of target groups, so this function was allocating a lot of memory
which then immediately became garbage. Re-using the structures across
an entire Sync saves effort.
Signed-off-by: Bryan Boreham <bjboreham@gmail.com>
Save work converting to `Labels` then to `Builder`.
`PopulateLabels()` now takes as Builder as input.
Signed-off-by: Bryan Boreham <bjboreham@gmail.com>
