promql: Improve histogram_quantile calculation for classic buckets
Tiny differences between classic buckets are most likely caused by floating point precision issues. With this commit, relative changes below a certain threshold are ignored. This makes the result of histogram_quantile more meaningful, and also avoids triggering the _input to histogram_quantile needed to be fixed for monotonicity_ annotations in unactionable cases.
This commit also adds explanation of the new adjustment and of the monotonicity annotation to the documentation of `histogram_quantile`.
---------
Signed-off-by: Jeanette Tan <jeanette.tan@grafana.com>
* Add benchmark for native histograms
This commit adds a PromQL benchmark for queries on native histograms.
Signed-off-by: Filip Petkovski <filip.petkovsky@gmail.com>
This PR adds an Experimental flag to the functions.
This can be used by https://github.com/prometheus/prometheus/pull/13059
but also xrate and other future functions.
Signed-off-by: Julien Pivotto <roidelapluie@o11y.eu>
This function is useful to analyze promQL queries. We want to use this in Mimir to record the time range which the query touches.
I also chose to remove the `Engine` receiver because it was unnecessary, and it makes it easier to use, but happy to refactor that if you disagree.
The function is untested on its own. If you prefer to have unit tests now that its exported, I can look into adding some.
Signed-off-by: Dimitar Dimitrov <dimitar.dimitrov@grafana.com>
* Remove NewPossibleNonCounterInfo until it can be made more efficient, and avoid creating empty annotations as much as possible
Signed-off-by: Jeanette Tan <jeanette.tan@grafana.com>
Return annotations (warnings and infos) from PromQL queries
This generalizes the warnings we have already used before (but only for problems with remote read) as "annotations".
Annotations can be warnings or infos (the latter could be false positives). We do not treat them different in the API for now and return them all as "warnings". It would be easy to distinguish them and return infos separately, should that appear useful in the future.
The new annotations are then used to create a lot of warnings or infos during PromQL evaluations. Partially these are things we have wanted for a long time (e.g. inform the user that they have applied `rate` to a metric that doesn't look like a counter), but the new native histograms have created even more needs for those annotations (e.g. if a query tries to aggregate float numbers with histograms).
The annotations added here are not yet complete. A prominent example would be a warning about a range too short for a rate calculation. But such a warnings is more tricky to create with good fidelity and we will tackle it later.
Another TODO is to take annotations into account when evaluating recording rules.
---------
Signed-off-by: Jeanette Tan <jeanette.tan@grafana.com>
Make it more likely that contributors will run the benchmark suite.
count_values needs more than 2GB at 1,000 steps, so just run it for 100.
And remove 10-step variant because it doesn't add much to 100 and
1000-step benchmarks.
Signed-off-by: Bryan Boreham <bjboreham@gmail.com>
Otherwise we have a highly unusual situation of over 100 chunks
in the headChunks list of each series, which heavily skews
performance.
Signed-off-by: Bryan Boreham <bjboreham@gmail.com>
promql: Extend testing framework to support native histograms
This includes both the internal testing framework as well as the rules unit test feature of promtool.
This also adds a bunch of basic tests. Many of the code level tests can now be converted to tests within the framework, and more tests can be added easily.
---------
Signed-off-by: Harold Dost <h.dost@criteo.com>
Signed-off-by: Gregor Zeitlinger <gregor.zeitlinger@grafana.com>
Signed-off-by: Stephen Lang <stephen.lang@grafana.com>
Co-authored-by: Harold Dost <h.dost@criteo.com>
Co-authored-by: Stephen Lang <stephen.lang@grafana.com>
Co-authored-by: Gregor Zeitlinger <gregor.zeitlinger@grafana.com>
So far, `ValidateHistogram` would not detect if the count did not
include the count in the zero bucket. This commit fixes the problem
and updates all the tests that have been undetected offenders so far.
Note that this problem would only ever create false negatives, so we
never falsely rejected to store a histogram because of it.
On the other hand, `ValidateFloatHistogram` has been to strict with
the count being at least as large as the sum of the counts in all the
buckets. Float precision issues could create false positives here, see
products of PromQL evaluations, it's actually quite hard to put an
upper limit no the floating point imprecision. Users could produce the
weirdest expressions, maxing out float precision problems. Therefore,
this commit simply removes that particular check from
`ValidateFloatHistogram`.
Signed-off-by: beorn7 <beorn@grafana.com>
promql engine: check unique labels using existing map
ContainsSameLabelset constructs a map with the same hash key as the one used to compile the output of rangeEval, so we can use that one and save work.
Need to hold the timestamp so we can be sure we saw the same series in the same evaluation.
`ContainsSameLabelset` constructs a map with the same hash key as
the one used to compile the output of `rangeEval`, so we can use that
one and save work.
Need to hold the timestamp so we can be sure we saw the same series
in the same evaluation.
Signed-off-by: Bryan Boreham <bjboreham@gmail.com>
The operator changes the meaning of the metric, so the metric name should
be dropped. Technically this would be a breaking change, but it's also very
obviously a bug and not likely that anyone depends on it.
Signed-off-by: Julius Volz <julius.volz@gmail.com>
Convert QueryOpts to an interface so that downstream projects like
https://github.com/thanos-community/promql-engine could extend the query
options with engine specific options that are not in the original
engine.
Will be used to enable query analysis per-query.
Signed-off-by: Giedrius Statkevičius <giedrius.statkevicius@vinted.com>
Handle more arithmetic operators and aggregators for native histograms
This includes operators for multiplication (formerly known as scaling), division, and subtraction. Plus aggregations for average and the avg_over_time function.
Stdvar and stddev will (for now) ignore histograms properly (rather than counting them but adding a 0 for them).
Signed-off-by: Jeanette Tan <jeanette.tan@grafana.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>
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>
In the past, every sample value was a float, so it was fine to call a
variable holding such a float "value" or "sample". With native
histograms, a sample might have a histogram value. And a histogram
value is still a value. Calling a float value just "value" or "sample"
or "V" is therefore misleading. Over the last few commits, I already
renamed many variables, but this cleans up a few more places where the
changes are more invasive.
Note that we do not to attempt naming in the JSON APIs or in the
protobufs. That would be quite a disruption. However, internally, we
can call variables as we want, and we should go with the option of
avoiding misunderstandings.
Signed-off-by: beorn7 <beorn@grafana.com>
In other words: Instead of having a “polymorphous” `Point` that can
either contain a float value or a histogram value, use an `FPoint` for
floats and an `HPoint` for histograms.
This seemingly small change has a _lot_ of repercussions throughout
the codebase.
The idea here is to avoid the increase in size of `Point` arrays that
happened after native histograms had been added.
The higher-level data structures (`Sample`, `Series`, etc.) are still
“polymorphous”. The same idea could be applied to them, but at each
step the trade-offs needed to be evaluated.
The idea with this change is to do the minimum necessary to get back
to pre-histogram performance for functions that do not touch
histograms. Here are comparisons for the `changes` function. The test
data doesn't include histograms yet. Ideally, there would be no change
in the benchmark result at all.
First runtime v2.39 compared to directly prior to this commit:
```
name old time/op new time/op delta
RangeQuery/expr=changes(a_one[1d]),steps=1-16 391µs ± 2% 542µs ± 1% +38.58% (p=0.000 n=9+8)
RangeQuery/expr=changes(a_one[1d]),steps=10-16 452µs ± 2% 617µs ± 2% +36.48% (p=0.000 n=10+10)
RangeQuery/expr=changes(a_one[1d]),steps=100-16 1.12ms ± 1% 1.36ms ± 2% +21.58% (p=0.000 n=8+10)
RangeQuery/expr=changes(a_one[1d]),steps=1000-16 7.83ms ± 1% 8.94ms ± 1% +14.21% (p=0.000 n=10+10)
RangeQuery/expr=changes(a_ten[1d]),steps=1-16 2.98ms ± 0% 3.30ms ± 1% +10.67% (p=0.000 n=9+10)
RangeQuery/expr=changes(a_ten[1d]),steps=10-16 3.66ms ± 1% 4.10ms ± 1% +11.82% (p=0.000 n=10+10)
RangeQuery/expr=changes(a_ten[1d]),steps=100-16 10.5ms ± 0% 11.8ms ± 1% +12.50% (p=0.000 n=8+10)
RangeQuery/expr=changes(a_ten[1d]),steps=1000-16 77.6ms ± 1% 87.4ms ± 1% +12.63% (p=0.000 n=9+9)
RangeQuery/expr=changes(a_hundred[1d]),steps=1-16 30.4ms ± 2% 32.8ms ± 1% +8.01% (p=0.000 n=10+10)
RangeQuery/expr=changes(a_hundred[1d]),steps=10-16 37.1ms ± 2% 40.6ms ± 2% +9.64% (p=0.000 n=10+10)
RangeQuery/expr=changes(a_hundred[1d]),steps=100-16 105ms ± 1% 117ms ± 1% +11.69% (p=0.000 n=10+10)
RangeQuery/expr=changes(a_hundred[1d]),steps=1000-16 783ms ± 3% 876ms ± 1% +11.83% (p=0.000 n=9+10)
```
And then runtime v2.39 compared to after this commit:
```
name old time/op new time/op delta
RangeQuery/expr=changes(a_one[1d]),steps=1-16 391µs ± 2% 547µs ± 1% +39.84% (p=0.000 n=9+8)
RangeQuery/expr=changes(a_one[1d]),steps=10-16 452µs ± 2% 616µs ± 2% +36.15% (p=0.000 n=10+10)
RangeQuery/expr=changes(a_one[1d]),steps=100-16 1.12ms ± 1% 1.26ms ± 1% +12.20% (p=0.000 n=8+10)
RangeQuery/expr=changes(a_one[1d]),steps=1000-16 7.83ms ± 1% 7.95ms ± 1% +1.59% (p=0.000 n=10+8)
RangeQuery/expr=changes(a_ten[1d]),steps=1-16 2.98ms ± 0% 3.38ms ± 2% +13.49% (p=0.000 n=9+10)
RangeQuery/expr=changes(a_ten[1d]),steps=10-16 3.66ms ± 1% 4.02ms ± 1% +9.80% (p=0.000 n=10+9)
RangeQuery/expr=changes(a_ten[1d]),steps=100-16 10.5ms ± 0% 10.8ms ± 1% +3.08% (p=0.000 n=8+10)
RangeQuery/expr=changes(a_ten[1d]),steps=1000-16 77.6ms ± 1% 78.1ms ± 1% +0.58% (p=0.035 n=9+10)
RangeQuery/expr=changes(a_hundred[1d]),steps=1-16 30.4ms ± 2% 33.5ms ± 4% +10.18% (p=0.000 n=10+10)
RangeQuery/expr=changes(a_hundred[1d]),steps=10-16 37.1ms ± 2% 40.0ms ± 1% +7.98% (p=0.000 n=10+10)
RangeQuery/expr=changes(a_hundred[1d]),steps=100-16 105ms ± 1% 107ms ± 1% +1.92% (p=0.000 n=10+10)
RangeQuery/expr=changes(a_hundred[1d]),steps=1000-16 783ms ± 3% 775ms ± 1% -1.02% (p=0.019 n=9+9)
```
In summary, the runtime doesn't really improve with this change for
queries with just a few steps. For queries with many steps, this
commit essentially reinstates the old performance. This is good
because the many-step queries are the one that matter most (longest
absolute runtime).
In terms of allocations, though, this commit doesn't make a dent at
all (numbers not shown). The reason is that most of the allocations
happen in the sampleRingIterator (in the storage package), which has
to be addressed in a separate commit.
Signed-off-by: beorn7 <beorn@grafana.com>
query_samples_total is a counter that tracks the total number of samples loaded by all queries.
The goal with this metric is to be able to see the amount of 'work' done by Prometheus to service queries.
At the moment we have metrics with the number of queries, plus more detailed metrics showing how much time each step of a query takes.
While those metrics do help they don't show us the whole picture.
Queries that do load more samples are (in general) more expensive than queries that do load fewer samples.
This means that looking only at the number of queries doesn't tell us how much 'work' Prometheus received.
Adding a counter that tracks the total number of samples loaded allows us to see if there was a spike in the cost of queries, not just the number of them.
Signed-off-by: Łukasz Mierzwa <l.mierzwa@gmail.com>
* Use zeropool.Pool to workaround SA6002
I built a tiny library called https://github.com/colega/zeropool to
workaround the SA6002 staticheck issue.
While searching for the references of that SA6002 staticheck issues on
Github first results was Prometheus itself, with quite a lot of ignores
of it.
This changes the usages of `sync.Pool` to `zeropool.Pool[T]` where a
pointer is not available.
Also added a benchmark for HeadAppender Append/Commit when series
already exist, which is one of the most usual cases IMO, as I didn't find
any.
Signed-off-by: Oleg Zaytsev <mail@olegzaytsev.com>
* Improve BenchmarkHeadAppender with more cases
Signed-off-by: Oleg Zaytsev <mail@olegzaytsev.com>
* A little copying is better than a little dependency
https://www.youtube.com/watch?v=PAAkCSZUG1c&t=9m28s
Signed-off-by: Oleg Zaytsev <mail@olegzaytsev.com>
* Fix imports order
Signed-off-by: Oleg Zaytsev <mail@olegzaytsev.com>
* Add license header
Signed-off-by: Oleg Zaytsev <mail@olegzaytsev.com>
* Copyright should be on one of the first 3 lines
Signed-off-by: Oleg Zaytsev <mail@olegzaytsev.com>
* Use require.Equal for testing
I don't depend on testify in my lib, but here we have it available.
Signed-off-by: Oleg Zaytsev <mail@olegzaytsev.com>
* Avoid flaky test
Signed-off-by: Oleg Zaytsev <mail@olegzaytsev.com>
* Also use zeropool for pointsPool in engine.go
Signed-off-by: Oleg Zaytsev <mail@olegzaytsev.com>
---------
Signed-off-by: Oleg Zaytsev <mail@olegzaytsev.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>
In Thanos we would like to start experimenting with custom functions that are
currently not part of the PromQL spec. We would do this by adding an implementation
for those functions in the Thanos engine: https://github.com/thanos-community/promql-engine and allow
users to decide which engine they want to use on a per-query basis.
Since we use the PromQL parser from Prometheus, injecting functions in the global `Functions` variable
would mean they also become available for the Prometheus engine. To avoid this side-effect, this commit
exposes a Parser interface in which the supported functions can be injected as an option. If not functions
are injected, the parser implementation will default to the functions defined in the global Functions variable.
Signed-off-by: Filip Petkovski <filip.petkovsky@gmail.com>
TestConcurrentRangeQueries runs many queries, up to 4 at the same time,
to try to expose any race conditions.
This change stops four of them from running with a thousand or more steps:
`holt_winters(a_X[1d], 0.3, 0.3)`
`changes(a_X[1d])`
`rate(a_X[1d])`
`absent_over_time(a_X[1d])`
Particularly when the test runs with `-race` in CI, this reduces the
time and resources required.
Signed-off-by: Bryan Boreham <bjboreham@gmail.com>
This is a bit more conservative than we could be. As long as a chunk
isn't the first in a block, we can be pretty sure that the previous
chunk won't disappear. However, the incremental gain of returning
NotCounterReset in these cases is probably very small and might not be
worth the code complications.
Wwith this, we now also pay attention to an explicitly set counter
reset during ingestion. While the case doesn't show up in practice
yet, there could be scenarios where the metric source knows there was
a counter reset even if it might not be visible from the values in the
histogram. It is also useful for testing.
Signed-off-by: beorn7 <beorn@grafana.com>
Alexander Trost