Commit graph

421 commits

Author SHA1 Message Date
Aliaksandr Valialkin
aa31cc85f5
app/vmselect/promql: properly return an empty result from limit_offset() if offset exceeds the number of inner time series
Updates https://github.com/VictoriaMetrics/VictoriaMetrics/issues/3312
2022-12-10 21:17:54 -08:00
Aliaksandr Valialkin
4eec43bf26
app/vmselect/promql: follow-up for 930f1ee153
Document the change at docs/CHANGELOG.md
Apply it to histogram_quantile() in the same way as to histogram_share()

Updates https://github.com/VictoriaMetrics/VictoriaMetrics/pull/3225
2022-11-07 14:46:00 +02:00
Siqi Liu
b8284e7ccb
BUGFIX: properly calculate histogram_quantile with the same value and different string le (#3225)
Co-authored-by: 647(siki.liu) <siki.liu@huolala.cn>
2022-11-07 14:45:58 +02:00
Aliaksandr Valialkin
0d6e30f3dc
app/vmselect/promql: expose missing metric vm_cache_size_max_bytes{type="promql/rollupResult"} 2022-11-07 14:29:45 +02:00
Aliaksandr Valialkin
cba8d34dab
app/vmselect/promql: properly calculate vm_rows_scanned_per_query histogram for rollup functions, which take into account only a few samples on the provided lookbehind window 2022-10-06 23:27:04 +03:00
Aliaksandr Valialkin
6c66848056
app/vmselect/promql: properly calculate quantiles_over_time() over a single raw sample 2022-10-06 22:38:18 +03:00
Roman Khavronenko
38568254c5
app/vmselect: ignore empty series for limit_offset (#3178)
* app/vmselect: ignore empty series for `limit_offset`

VictoriaMetrics doesn't return empty series (with all NaN values) to
the user. But such series are filtered after transform functions.
It means `limit_offset` will account for empty series as well.

For example, let's consider following data set:
```
time series:
foo{label="1"} NaN, NaN, NaN, NaN // empty series
foo{label="2"} 1, 2, 3, 4
foo{label="3"} 4, 3, 2, 1
```

When user requests all series for metric `foo` the empty series
will be filtered out:
```
/query=foo:
foo{label="v2"} 1, 2, 3, 4
foo{label="v3"} 4, 3, 2, 1
```

But `limit_offset(1, 1, foo)` is applied to original series, not filtered yet.
So it will return `foo{label="v2"}` (skips the first in list)
```
/query=limit_offset(1, 1, foo):
foo{label="v2"} 1, 2, 3, 4
```

Expected result would be to apply `limit_offset` to already filtered list,
so in result we receive `foo{label="v3"}`:
```
/query=limit_offset(1, 1, foo):
foo{label="v3"} 4, 3, 2, 1
```

The change does exactly that - filters empty series before applying `limit_offset`.

Signed-off-by: hagen1778 <roman@victoriametrics.com>

* app/vmselect: ignore empty series for `limit_offset`

Signed-off-by: hagen1778 <roman@victoriametrics.com>

Signed-off-by: hagen1778 <roman@victoriametrics.com>
2022-10-06 15:49:56 +03:00
Aliaksandr Valialkin
d385e1c60c
app/vmselect/promql: remove empty series before applying aggregate function
Previously empty series (e.g. series with all NaN samples) were passed to aggregate functions.
Such series must be ingored by all the aggregate functions.
So it is better from consistency PoV filtering out empty series before applying aggregate functions.
2022-10-06 15:49:55 +03:00
Roman Khavronenko
be67c5e27f
vmselect/rollup: rm workaround for slow-changing counters (#3163)
The workaround was introduced to fix https://github.com/VictoriaMetrics/VictoriaMetrics/issues/962.
However, it didn't prove itself useful. Instead, it is recommended using `increase_pure` function.

Removing the workaround makes VM to produce accurate results when calculating
`delta` or `increase` functions over slow-changing counters with vary intervals
between data points.

Signed-off-by: hagen1778 <roman@victoriametrics.com>

Signed-off-by: hagen1778 <roman@victoriametrics.com>
2022-10-06 15:46:29 +03:00
Aliaksandr Valialkin
727255085b
app/vmselect/promql: consistently calculate rate_over_sum(m[d]) as sum_over_time(m[d])/d
Updates https://github.com/VictoriaMetrics/VictoriaMetrics/issues/3045
2022-10-06 15:44:19 +03:00
Aliaksandr Valialkin
889c4d5261
app/vmselect/promql: typo fix after 992f36702f 2022-08-30 12:04:51 +03:00
Aliaksandr Valialkin
00af0141d5
app/vmselect/promql: follow-up after 2d71b4859c
- Use getScalar() function for obtaining the expected scalar from phi arg
- Reduce the error message returned to the user when incorrect phi is passed to histogram_quantiles
- Improve the description of this bugfix in the docs/CHANGELOG.md

Updates https://github.com/VictoriaMetrics/VictoriaMetrics/issues/3026
2022-08-30 11:21:33 +03:00
Dmytro Kozlov
968ca13d2e
vmselect/promql: fix panic in histogram_quantiles function (#3029)
* vmselect/promql: fix panic in histogram_quantiles function

* Update docs/MetricsQL.md

Co-authored-by: Roman Khavronenko <roman@victoriametrics.com>

Co-authored-by: Aliaksandr Valialkin <valyala@victoriametrics.com>
Co-authored-by: Roman Khavronenko <roman@victoriametrics.com>
2022-08-30 11:21:33 +03:00
Aliaksandr Valialkin
3d10ef084b
app/vmselect/promql/transform.go: reuse evalNumber() function for constructing timezone_offset() results 2022-08-08 16:39:09 +03:00
Aliaksandr Valialkin
708f1e8048
app/vmselect/promql: properly return q1 series from q1 ifnot q2 when q2 returns nothing 2022-08-08 15:44:51 +03:00
Aliaksandr Valialkin
919939ee9f
app/vmselect/promql: consistency update after 93fbd0c54b 2022-07-13 12:34:28 +03:00
Roman Khavronenko
214e01499c
promql: return step as scrapeInterval when it can't be calculated (#2865)
The change allows to specify default value for `getScrapeInterval`
function when actual interval can't be calculated.

Before the change, function were returning `maxSilenceInterval` (5m)
in such cases, which may be not correct for instant queries processing.
The specific scenario where using `maxSilenceInterval` caused issues
is the following:
1. Series becomes stale;
2. Client (in this case vmalert) continues to request series every 15s;
3. Database returns empty results as expected;
4. But at some specific moment of time database returns datapoints from `now()-5m`,
because lookback window was extended to `maxSilenceInterval`.

Signed-off-by: hagen1778 <roman@victoriametrics.com>
2022-07-13 12:34:28 +03:00
Aliaksandr Valialkin
5867708b3d
app/vmselect/promql: validate function name before evaluating its arguments
This avoids unneeded evaluation of args for unknown functions
2022-07-12 19:49:54 +03:00
Roman Khavronenko
78da074ad3
metricsql: properly evaluate timezone_offset over time interval (#2842)
* metricsql: properly evaluate `timezone_offset` over time interval

https://github.com/VictoriaMetrics/VictoriaMetrics/issues/2771
Signed-off-by: hagen1778 <roman@victoriametrics.com>

* Update docs/CHANGELOG.md

Co-authored-by: Aliaksandr Valialkin <valyala@victoriametrics.com>
2022-07-08 14:04:40 +03:00
Aliaksandr Valialkin
f4df43f7cc
app/vmselect/netstorage: remove unused auth.Token arg 2022-07-06 12:46:21 +03:00
Aliaksandr Valialkin
10c63f5853
app/vmselect: make fmt after f3ece83e67 2022-07-05 14:56:53 +03:00
Aliaksandr Valialkin
8516670582
app/vmselect/promql: properly calculate histogram_quantile over unexpected le buckets
See https://github.com/VictoriaMetrics/VictoriaMetrics/issues/2819
2022-07-05 13:20:41 +03:00
Aliaksandr Valialkin
fa08220d27
app/vmselect/promql: properly handle partial counter resets in rate(), irate(), increase() and remove_resets() functions
Updates https://github.com/VictoriaMetrics/VictoriaMetrics/issues/2787
2022-06-30 22:40:26 +03:00
Aliaksandr Valialkin
4fb0f15322
all: readability improvements for query traces
- show dates in human-readable format, e.g. 2022-05-07, instead of a numeric value
- limit the maximum length of queries and filters shown in trace messages
2022-06-30 18:19:43 +03:00
Aliaksandr Valialkin
daefb64f38
app/vmselect: expose additional histograms at /metrics page, which may help get more insights for the query workload
This commit is based on https://github.com/VictoriaMetrics/VictoriaMetrics/pull/2792
2022-06-28 20:18:31 +03:00
Aliaksandr Valialkin
9adff0b686
app/vmselect/promql: show the number of scanned samples in the query trace 2022-06-28 19:32:03 +03:00
Aliaksandr Valialkin
399d4c36ae
app/vmselect: optimize /api/v1/series a bit for time ranges smaller than one day 2022-06-28 12:55:20 +03:00
Aliaksandr Valialkin
6386f117c8
all: show timeRange in traces in human-readable format instead of timestamps in milliseconds 2022-06-27 13:42:57 +03:00
Aliaksandr Valialkin
fe2269b999
all: remove explicit "xxhash" name when importing github.com/cespare/xxhash/v2 package
This package already has the same name, so there is no need in explicit name
2022-06-21 20:24:28 +03:00
Aliaksandr Valialkin
a9ea3fee38
lib/querytracer: make it easier to use by passing trace context message to New and NewChild
The context message can be extended by calling Donef.
If there is no need to extend the message, then just call Done.
2022-06-08 21:16:12 +03:00
Aliaksandr Valialkin
afced37c0b
all: add initial support for query tracing
See https://docs.victoriametrics.com/Single-server-VictoriaMetrics.html#query-tracing

Updates https://github.com/VictoriaMetrics/VictoriaMetrics/issues/1403
2022-06-01 02:31:44 +03:00
Aliaksandr Valialkin
68bfa05437
app/vmselect/promql: do not return values from label_value() if the original time series has no values at the selected timestamps 2022-05-09 17:58:39 +03:00
Aliaksandr Valialkin
6eb1580158
app/vmselect/promql: add tlast_change_over_time(m[d]) function, which returns the timestamp for the last change of m on the given lookbehind window d 2022-04-27 10:58:40 +03:00
Aliaksandr Valialkin
dac24aa342
app/vmselect/promql: properly handle scalar default vector, scalar if vector and scalar ifnot vector queries
Previously `vector` time series could be unexpectedly returned from such queries
2022-04-21 15:34:14 +03:00
Aliaksandr Valialkin
bc54ae9608
app/vmselect/promql: fix comparison to nan
The comparison to nan has been broken in d335cc886c

Updates https://github.com/VictoriaMetrics/VictoriaMetrics/issues/150
2022-04-21 14:56:41 +03:00
Aliaksandr Valialkin
a5cfe5d13e
app/vmselect/promql: add drop_common_labels() function 2022-04-21 14:20:36 +03:00
Aliaksandr Valialkin
7ca08945fd
app/vmselect/promql: fix q default b where b may have empty time series 2022-04-21 11:43:09 +03:00
Aliaksandr Valialkin
25b841c6ed
app/vmselect/promql: fix duplicate time series error on joins against time series filtered by values
This should prevent from `duplicate time series` errors when executing the following query:

kube_pod_container_resource_requests{resource="cpu"} * on (namespace,pod) group_left() (kube_pod_status_phase{phase=~"Pending|Running"}==1)

where `kube_pod_status_phase{phase=~"Pending|Running"}==1` filters out diplicate time series
2022-04-20 22:21:20 +03:00
Aliaksandr Valialkin
0ef7a05fc0
app/vmselect/promql: rename removeNaNs() to more clear removeEmptySeries() 2022-04-20 19:53:24 +03:00
Aliaksandr Valialkin
3605e1743b
app/vmselect/promql: allow calling InitRollupResultCache+StopRollupResultCache multiple times during tests 2022-04-11 12:32:23 +03:00
Aliaksandr Valialkin
b843f0e229
app/vmselect: add fine-grained limits for the number of returned/scanned time series for various APIs 2022-03-26 11:28:14 +02:00
Aliaksandr Valialkin
99ee1c13c0
app/vmselect/promql: reduce the maximum number of label values, which can be propagated from one side of the binary operation to another side of the binary operation from 10K to 1K
There are user reports that 10K unique values in a single label filter may lead to performance and memory usage issues
2022-02-24 04:03:28 +02:00
Aliaksandr Valialkin
244c23ea2c
lib/workingsetcache: reduce the default cache rotation period from hour to 20 minutes
This should reduce memory usage under high time series churn rate
2022-02-23 13:42:27 +02:00
Aliaksandr Valialkin
989668beba
app/vmselect/promql: return at most one time series from absent_over_time() in the same way as Prometheus does
See https://github.com/VictoriaMetrics/VictoriaMetrics/issues/2130
2022-02-12 15:46:43 +02:00
Aliaksandr Valialkin
1173964d8d
app/vmselect/promql: clarify comments on why the right side of if and and operators are executed at first 2022-02-03 00:26:42 +02:00
Aliaksandr Valialkin
10367d7e4c
app/vmselect/promql: do not push down filters, which enumerate more than 10k unique values
Such filters may slow down time series search, so just skip them.

This is a follow-up for e7f1ceeb84

Updates https://github.com/VictoriaMetrics/VictoriaMetrics/issues/1827
2022-02-02 23:42:25 +02:00
Aliaksandr Valialkin
13c2692ca1
app/vmselect/promql: properly handle foo or bar queries
Such queries may miss `bar` results after the commit e7f1ceeb84
because common label filters from `foo` could be mistakenly applied to `bar`.
2022-02-01 17:43:10 +02:00
Aliaksandr Valialkin
02b2bfcff3
lib/bytesutil: split Resize* funcs to MayOverallocate and NoOverallocate for more fine-grained control over memory allocations
Follow-up for f4989edd96

Updates https://github.com/VictoriaMetrics/VictoriaMetrics/issues/2007
2022-02-01 11:20:20 +02:00
Aliaksandr Valialkin
784cdcb080
app/vmselect/promql: check for binary operation in case-insensitive manner when deciding which side of the operation to perform the first
PromQL and MetricsQL operators are case-insensitive
2022-01-31 22:07:35 +02:00
Aliaksandr Valialkin
04d6596298
app/vmselect/promql: optimize queries, which join on _info metrics.
Automatically add common filters from one side of binary operation
to the other side before sending the query to storage subsystem.

See https://grafana.com/blog/2021/08/04/how-to-use-promql-joins-for-more-effective-queries-of-prometheus-metrics-at-scale/
and https://www.robustperception.io/exposing-the-software-version-to-prometheus
2022-01-31 20:25:15 +02:00