Two issues are fixed here, that lead to the same problem:
1. If `newSampleRing` is called with an unknown ValueType including
ValueNone, we have initialized the interface buffer (`iBuf`).
However, we would still use a specialized buffer for the first
sample, opportunistically assuming that we might still not
encounter mixed samples and we should go down the more efficient
road.
2. If the `sampleRing` is `reset`, we leave all buffers alone,
including `iBuf`, which is generally fine, but not for `iBuf`, see
below.
In both cases, `iBuf` already contains values, but we will fill one of
the specialized buffers first. Once we then actually encounter mixed
samples, the content of the specialized buffer is copied into `iBuf`
using `append`. That's by itself the right idea because `iBuf` might
be `nil`, and even if not, it might or might not have the right
capacity. However, this approach assumes that `iBuf` is empty, or more
precisely has a length of zero.
This commit makes sure that `iBuf` does not get needlessly initialized
in `newSampleRing` and that it is emptied upon `reset`.
A test case is added to demonstrate both issues above.
Signed-off-by: beorn7 <beorn@grafana.com>
Instead of setting to nil and allocating a new slice every time the
merge is advanced, re-use the previous slice.
This is safe because the `currentSets` member is only used inside member
functions, and explicitly copied in `At()`, the only place it leaves the
struct.
Signed-off-by: Bryan Boreham <bjboreham@gmail.com>
The new version features a set of test cases that simplify the addition
of new HTTP status codes.
Signed-off-by: William Dumont <william.dumont@grafana.com>
Header name is `Retry-Attempt`, only set when >0.
Signed-off-by: Marc Tuduri <marctc@protonmail.com>
Signed-off-by: Paschalis Tsilias <paschalis.tsilias@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>
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>
Add a chunk size limit in bytes
This creates a hard cap for XOR chunks of 1024 bytes.
The limit for histogram chunk is also 1024 bytes, but it is a soft limit as a histogram has a dynamic size, and even a single one could be larger than 1024 bytes.
This also avoids cutting new histogram chunks if the existing chunk has fewer than 10 histograms yet. In that way, we are accepting "jumbo chunks" in order to have at least 10 histograms in a chunk, allowing compression to kick in.
Signed-off-by: Justin Lei <justin.lei@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>
* Add OTLP Ingestion endpoint
We copy files from the otel-collector-contrib. See the README in
`storage/remote/otlptranslator/README.md`.
This supersedes: https://github.com/prometheus/prometheus/pull/11965
Signed-off-by: gouthamve <gouthamve@gmail.com>
* Return a 200 OK
It is what the OTEL Golang SDK expect :(
https://github.com/open-telemetry/opentelemetry-go/issues/4363
Signed-off-by: Goutham <gouthamve@gmail.com>
---------
Signed-off-by: gouthamve <gouthamve@gmail.com>
Signed-off-by: Goutham <gouthamve@gmail.com>
* WIP implement WAL watcher reading via notifications over a channel from
the TSDB code
Signed-off-by: Callum Styan <callumstyan@gmail.com>
* Notify via head appenders Commit (finished all WAL logging) rather than
on each WAL Log call
Signed-off-by: Callum Styan <callumstyan@gmail.com>
* Fix misspelled Notify plus add a metric for dropped Write notifications
Signed-off-by: Callum Styan <callumstyan@gmail.com>
* Update tests to handle new notification pattern
Signed-off-by: Callum Styan <callumstyan@gmail.com>
* this test maybe needs more time on windows?
Signed-off-by: Callum Styan <callumstyan@gmail.com>
* does this test need more time on windows as well?
Signed-off-by: Callum Styan <callumstyan@gmail.com>
* read timeout is already a time.Duration
Signed-off-by: Callum Styan <callumstyan@gmail.com>
* remove mistakenly commited benchmark data files
Signed-off-by: Callum Styan <callumstyan@gmail.com>
* address some review feedback
Signed-off-by: Callum Styan <callumstyan@gmail.com>
* fix missed changes from previous commit
Signed-off-by: Callum Styan <callumstyan@gmail.com>
* Fix issues from wrapper function
Signed-off-by: Callum Styan <callumstyan@gmail.com>
* try fixing race condition in test by allowing tests to overwrite the
read ticker timeout instead of calling the Notify function
Signed-off-by: Callum Styan <callumstyan@gmail.com>
* fix linting
Signed-off-by: Callum Styan <callumstyan@gmail.com>
---------
Signed-off-by: Callum Styan <callumstyan@gmail.com>
The storage.ChunkSeries iterator assumes that a histogram sample can always be
appended to the currently open chunk. This is not the case when there is a counter reset,
or when appending a stale sample to a chunk with non-stale samples. In addition, the open chunk sometimes
needs to be recoded before a sample can be appended.
This commit addresses the issue by implementing a RecodingAppender which can recode incoming
samples in a transparent way. It also detects cases when a sample cannot be appended at all and
returns `false` so that the caller can open a new chunk.
Signed-off-by: Filip Petkovski <filip.petkovsky@gmail.com>
Signed-off-by: György Krajcsovits <gyorgy.krajcsovits@grafana.com>
Signed-off-by: Ganesh Vernekar <ganeshvern@gmail.com>
Co-authored-by: György Krajcsovits <gyorgy.krajcsovits@grafana.com>
Co-authored-by: Ganesh Vernekar <ganeshvern@gmail.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>
Signed-off-by: Sebastian Rabenhorst <sebastian.rabenhorst@shopify.com>
Fixed sampleRingIterator for mixed histograms
Signed-off-by: Sebastian Rabenhorst <sebastian.rabenhorst@shopify.com>
Fixed lint
This commit is doing what I would have expected that Go generics do
for me. However, the PromQL benchmarks show a significant runtime and
allocation increase with `genericAdd`, so this replaces it with
hand-coded non-generic versions of it.
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>
This utilizes the fact that most sampleRings will only contain samples
of one type. In this case, the generic interface is circumvented, and
a bespoke buffer for the one actually occurring sample type is
used. Should a sampleRing receive a sample of a different kind later,
it will transparently switch to the generic behavior.
Signed-off-by: beorn7 <beorn@grafana.com>
Previously, we had one “polymorphous” `sample` type in the `storage`
package. This commit breaks it up into `fSample`, `hSample`, and
`fhSample`, each still implementing the `tsdbutil.Sample` interface.
This reduces allocations in `sampleRing.Add` but inflicts the penalty
of the interface wrapper, which makes things worse in total.
This commit therefore just demonstrates the step taken. The next
commit will tackle the interface overhead problem.
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>
During remote write, we call url.String() twice:
- to add the Endpoint() to the span
- to actually know where whe should send the request
This value does not change over time, and it's not really that
lightweight to calculate. I wrote this simple benchmark:
func BenchmarkURLString(b *testing.B) {
u, err := url.Parse("https://remote.write.com/api/v1")
require.NoError(b, err)
b.Run("string", func(b *testing.B) {
count := 0
for i := 0; i < b.N; i++ {
count += len(u.String())
}
})
}
And the results are ~200ns/op, 80B/op, 3 allocs/op.
Yes, we're going to go to the network here, which is a huge amount of
resources compared to this, but still, on agents that send 500 requests
per second, that is 1500 wasteful allocations per second.
Signed-off-by: Oleg Zaytsev <mail@olegzaytsev.com>
In storage/remote, try converting to RecoverableError using errors.As,
instead of through direct casting.
Signed-off-by: Arve Knudsen <arve.knudsen@gmail.com>
As far as I understand it, we'd never expect to receive a nil span,
and remote.spansProtoToSpans would panic if we received a nil span.
Marking the fields as non-nullable also means the generated Golang
code doesn't use pointers for these fields, reducing allocations.
Signed-off-by: Charles Korn <charles.korn@grafana.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>
This is an optimization on the existing append in OOOChunk.
What we've been doing so far is find the place inside the out-of-order
slice where the new sample should go in and then place it there and move
any samples to the right if necessary. This is OK but requires a binary
search every time the slice is bigger than 0.
The optimization is opinionated and suggests that although out-of-order
samples can be out-of-order amongst themselves they'll probably be in
order thus we can probably optimistically append at the end and if not
do the binary search.
OOOChunks are capped to 30 samples by default so this is a small
optimization but everything adds up, specially if you handle many active
timeseries with out-of-order samples.
Signed-off-by: Jesus Vazquez <jesus.vazquez@grafana.com>
Signed-off-by: Jesus Vazquez <jesusvazquez@users.noreply.github.com>
Co-authored-by: Ganesh Vernekar <ganeshvern@gmail.com>
* adapt code.go and write_handler.go to support float histograms
* adapt watcher.go to support float histograms
* wip adapt queue_manager.go to support float histograms
* address comments for metrics in queue_manager.go
* set test cases for queue manager
* use same counts for histograms and float histograms
* refactor createHistograms tests
* fix float histograms ref in watcher_test.go
* address PR comments
Signed-off-by: Marc Tuduri <marctc@protonmail.com>
In most cases, there is no sample at `maxt`, so `PeekBack` has to be
used. So far, `PeekBack` did not return a float histogram, and we
disregarded even any returned normal histogram. This fixes both, and
also tweaks the unit test to discover the problem (by using an earlier
timestamp than "now" for the samples in the TSDB).
Signed-off-by: beorn7 <beorn@grafana.com>
Extends Appender.AppendHistogram function to accept the FloatHistogram. TSDB supports appending, querying, WAL replay, for this new type of histogram.
Signed-off-by: Marc Tudurí <marctc@protonmail.com>
Signed-off-by: Ganesh Vernekar <ganeshvern@gmail.com>
Co-authored-by: Ganesh Vernekar <ganeshvern@gmail.com>
`QueueManager.externalLabels` becomes a slice rather than a `Labels` so
we can index into it when doing the merge operation.
Note we avoid calling `Labels.Len()` in `labelProtosToLabels()`.
It isn't necessary - `append()` will enlarge the buffer and we're
expecting to re-use it many times.
Also, we now validate protobuf input before converting to Labels.
This way we can detect errors first, and we don't place unnecessary
requirements on the Labels structure.
Re-do seriesFilter using labels.Builder (albeit N^2).
Signed-off-by: Bryan Boreham <bjboreham@gmail.com>
This is required to preserve the interface property of SeriesSet that says
"At returns full series. Returned series should be iterable even after Next is called."
Signed-off-by: sniper91 <kevinzhao91@outlook.com>
Re-use previous memory if it is already of the correct type.
In `NewListSeries` we hoist the conversion to an interface value out
so it only allocates once.
Signed-off-by: Bryan Boreham <bjboreham@gmail.com>
Patterned after `Chunk.Iterator()`: pass the old iterator in so it
can be re-used to avoid allocating a new object.
(This commit does not do any re-use; it is just changing all the method
signatures so re-use is possible in later commits.)
Signed-off-by: Bryan Boreham <bjboreham@gmail.com>
The SeriesSets to be merged must be created each time round the loop,
otherwise the benchmark is not doing any real work.
Don't call ExpandSeries, because it spends most of its time allocating
a memory buffer to hold the result, which we don't look at.
Signed-off-by: Bryan Boreham <bjboreham@gmail.com>
Fix up merge test again
errors.Unwrap() actually dangerously returns nil if the error does not have an
Unwrap() method, which is the case in at least one of these places where I
noticed that no error was being logged at all when it should have.
Signed-off-by: Julius Volz <julius.volz@gmail.com>
* remote/read_handler: pool input to Marshal()
Use a sync.Pool to reuse byte slices between calls to Marshal() in the
remote read handler.
Signed-off-by: Giedrius Statkevičius <giedrius.statkevicius@vinted.com>
* remote: add microbenchmark for remote read handler
Signed-off-by: Giedrius Statkevičius <giedrius.statkevicius@vinted.com>
Signed-off-by: Giedrius Statkevičius <giedrius.statkevicius@vinted.com>
The wlog.WL type can now be used to create a Write Ahead Log or a Write
Behind Log.
Before the prefix for wbl metrics was
'prometheus_tsdb_out_of_order_wal_' and has been replaced with
'prometheus_tsdb_out_of_order_wbl_'.
Signed-off-by: Jesus Vazquez <jesus.vazquez@grafana.com>
Signed-off-by: Jesus Vazquez <jesusvazquez@users.noreply.github.com>
Co-authored-by: Ganesh Vernekar <15064823+codesome@users.noreply.github.com>
Where the code was multiplying bytes by number of operations, this
resulted in absurdly high throughput numbers.
Also, in `BenchmarkParse()`, don't run the `expfmt` case twice.
Signed-off-by: Bryan Boreham <bjboreham@gmail.com>