fix(logql): Make `LabelSampleExtractor` work as expected when label doesn't exist in logline #6766

kavirajk · 2022-07-25T22:54:31Z

What this PR does / why we need it:
Consider Log lines

msg=hello subqueries=5
msg=hello subqueries=0
msg=hello

If you run following logql

{foo="bar"}|logfmt|subqueries!=0

it returns

msg=hello subqueries=5

But if you run following logql

{foo="bar"}|logfmt|subqueries!="0" # NOTE: "0" as string.

it returns

msg=hello subqueries=5
msg=hello

NOTE: it also returns log lines that doesn't contain subqueries label in the first
place.

This cause subtle error if used in metric queries using that label in unwrap.

e.g: following logql on above log lines returns

sum_over_time({foo="bar"}|logfmt|subqueries!="0"|unwrap subqueries[1m])

Returns pipeline error: 'SampleExtractionErr'. Because, the lines without subqueries
labels are not skipped during ProcessLine and during metric extraction, our extractor
sets this SampleExtractionerror.

EDIT: After some discussions with @cyriltovena and @owen-d. We have decided fix only the sample extractor without changing any behavior of label filter.

Which issue(s) this PR fixes:
Fixes # #6713

Special notes for your reviewer:

Checklist

Documentation added
Tests updated
Is this an important fix or new feature? Add an entry in the CHANGELOG.md.
Changes that require user attention or interaction to upgrade are documented in docs/sources/upgrading/_index.md

kavirajk · 2022-07-25T22:54:59Z

pkg/logql/log/label_filter.go

@@ -342,7 +342,10 @@ func (s *StringLabelFilter) Process(_ int64, line []byte, lbs *LabelsBuilder) ([
 	if s.Name == logqlmodel.ErrorLabel {
 		return line, s.Matches(lbs.GetErr())
 	}
-	v, _ := lbs.Get(s.Name)
+	v, ok := lbs.Get(s.Name)


this is the real fix.

Consider Log lines ``` msg=hello subqueries=5 msg=hello subqueries=0 msg=hello ``` If you run following logql ``` {foo="bar"}|logfmt|subqueries!=0 ``` it returns ``` msg=hello subqueries=5 ``` But if you run following logql ``` {foo="bar"}|logfmt|subqueries!="0" # NOTE: "0" as string. ``` it returns ``` msg=hello subqueries=5 msg=hello ``` NOTE: it also returns log lines that doesn't contain `subqueries` label in the first place. This cause subtle error if used in metric queries using that `label` in unwrap. e.g: following logql on above log lines returns ``` sum_over_time({foo="bar"}|logfmt|subqueries!="0"|unwrap subqueries[1m]) ``` Returns `pipeline error: 'SampleExtractionErr'`. Because, the lines without `subqueries` labels are not skipped during `ProcessLine` and during metric extraction, our extractor sets this `SampleExtractionerror`. Fixes: #6713

Signed-off-by: Kaviraj <[email protected]>

grafanabot · 2022-07-25T23:10:24Z

./tools/diff_coverage.sh ../loki-main/test_results.txt test_results.txt ingester,distributor,querier,querier/queryrange,iter,storage,chunkenc,logql,loki

Change in test coverage per package. Green indicates 0 or positive change, red indicates that test coverage for a package fell.

+           ingester	0%
+        distributor	0%
+            querier	0%
+ querier/queryrange	0%
+               iter	0%
+            storage	0%
+           chunkenc	0%
+              logql	0%
+               loki	0%

ssncferreira

I'm a bit confused here because this seems related to the value type that we use in the query, i.e. either string != "0" or integer != 0.
If a log line does not contain a label, e.g.

msg=hello

I would always assume the value of the label subqueries would be the empty string "", and, therefore, it would be different than both string "0" and integer 0.
With the changes from this PR it seems this assumption is wrong, so it should be clearly documented

pkg/logql/log/label_filter_test.go

pkg/logql/log/metrics_extraction_test.go

Signed-off-by: Kaviraj <[email protected]>

grafanabot · 2022-07-27T08:45:56Z

./tools/diff_coverage.sh ../loki-main/test_results.txt test_results.txt ingester,distributor,querier,querier/queryrange,iter,storage,chunkenc,logql,loki

Change in test coverage per package. Green indicates 0 or positive change, red indicates that test coverage for a package fell.

+           ingester	0%
+        distributor	0%
+            querier	0%
+ querier/queryrange	0%
+               iter	0%
+            storage	0%
+           chunkenc	0%
+              logql	0%
-               loki	-0.6%

Signed-off-by: Kaviraj <[email protected]>

grafanabot · 2022-07-27T09:00:59Z

./tools/diff_coverage.sh ../loki-main/test_results.txt test_results.txt ingester,distributor,querier,querier/queryrange,iter,storage,chunkenc,logql,loki

Change in test coverage per package. Green indicates 0 or positive change, red indicates that test coverage for a package fell.

+           ingester	0%
+        distributor	0%
+            querier	0%
+ querier/queryrange	0%
+               iter	0%
+            storage	0%
+           chunkenc	0%
+              logql	0%
+               loki	0%

ssncferreira

LGTM! Thanks 🙏
Do you think it would also make sense to update the documentation with some examples? (can be part of a separate PR)

cyriltovena · 2022-07-28T07:40:39Z

pkg/logql/log/label_filter.go

+	// 4.msg="hello" subqueries=""
+	//```
+
+	// Now when quering like `{...}|logfmt|subqueries=""` should consider lines 3 and 4.


Not sure about this, it doesn't really make it easier to understand for the end user as this make it more obscure:

when using equality we consider non existent label empty

but for inequality we don't.

Honestly in #6713 I think the expectations were wrong.

sum(sum_over_time({job="loki-ops/query-frontend"} |= "metrics.go" | logfmt | subqueries!="0" | unwrap subqueries [5m] ))

When subqueries doesn't exist then subqueries!="0" is true.

What is weird was that empty labels does transform into a 0 integer.

When subqueries doesn't exist then subqueries!="0" is true.

So you saying subqueries!=0 and subqueries!="0" ok to return different log lines?. But that's unexpected behaviour to end-user IMHO. Both should return exactly same results IMO.

What is weird was that empty labels does transform into a 0 integer.

That's because, we don't mutate the log line if label is not present. So treat it as if label="" during label filter, But still log line doesn't contain actual label, which breaks when doing unwrap on same label during metrics extraction.

What is weird was that empty labels does transform into a 0 integer.

I agree with @cyriltovena. This is the part I think we need to address.

May be we discard empty values ? If that is possible.

Yes that's possible. We can skip empty values during metrics extraction when using unwrap without changing any of the existing label filter behavior.

Also the more I think about it, subqueries!="0" and subqueries!=0 are two different label filters and should be treated as such. And former can indeed match log lines that doesn't have subqueries label at all (we are assuming it has default empty string substring=" when doesn't exist).

I have change the PR like it's mentioned here.

Signed-off-by: Kaviraj <[email protected]>

grafanabot · 2022-08-01T13:51:29Z

./tools/diff_coverage.sh ../loki-main/test_results.txt test_results.txt ingester,distributor,querier,querier/queryrange,iter,storage,chunkenc,logql,loki

Change in test coverage per package. Green indicates 0 or positive change, red indicates that test coverage for a package fell.

+           ingester	0%
+        distributor	0%
+            querier	0%
+ querier/queryrange	0%
+               iter	0%
+            storage	0%
+           chunkenc	0%
+              logql	0%
+               loki	0%

grafanabot · 2022-08-01T13:56:56Z

./tools/diff_coverage.sh ../loki-main/test_results.txt test_results.txt ingester,distributor,querier,querier/queryrange,iter,storage,chunkenc,logql,loki

Change in test coverage per package. Green indicates 0 or positive change, red indicates that test coverage for a package fell.

+           ingester	0%
+        distributor	0%
+            querier	0%
+ querier/queryrange	0%
+               iter	0%
+            storage	0%
+           chunkenc	0%
+              logql	0%
+               loki	0%

grafanabot · 2022-08-01T14:02:02Z

./tools/diff_coverage.sh ../loki-main/test_results.txt test_results.txt ingester,distributor,querier,querier/queryrange,iter,storage,chunkenc,logql,loki

Change in test coverage per package. Green indicates 0 or positive change, red indicates that test coverage for a package fell.

+           ingester	0%
+        distributor	0%
+            querier	0%
+ querier/queryrange	0%
+               iter	0%
+            storage	0%
+           chunkenc	0%
+              logql	0%
+               loki	0%

cyriltovena

LGTM

I think the changelog needs a better documentation of the change.

Signed-off-by: Kaviraj <[email protected]>

grafanabot · 2022-08-03T07:34:49Z

./tools/diff_coverage.sh ../loki-main/test_results.txt test_results.txt ingester,distributor,querier,querier/queryrange,iter,storage,chunkenc,logql,loki

Change in test coverage per package. Green indicates 0 or positive change, red indicates that test coverage for a package fell.

+           ingester	0.1%
+        distributor	0%
+            querier	0%
+ querier/queryrange	0%
+               iter	0%
+            storage	0%
+           chunkenc	0%
+              logql	0%
+               loki	0%

…oesn't exist in logline (grafana#6766) * fix(logql): Make `StringLabelfilter` work as expected Consider Log lines ``` msg=hello subqueries=5 msg=hello subqueries=0 msg=hello ``` If you run following logql ``` {foo="bar"}|logfmt|subqueries!=0 ``` it returns ``` msg=hello subqueries=5 ``` But if you run following logql ``` {foo="bar"}|logfmt|subqueries!="0" # NOTE: "0" as string. ``` it returns ``` msg=hello subqueries=5 msg=hello ``` NOTE: it also returns log lines that doesn't contain `subqueries` label in the first place. This cause subtle error if used in metric queries using that `label` in unwrap. e.g: following logql on above log lines returns ``` sum_over_time({foo="bar"}|logfmt|subqueries!="0"|unwrap subqueries[1m]) ``` Returns `pipeline error: 'SampleExtractionErr'`. Because, the lines without `subqueries` labels are not skipped during `ProcessLine` and during metric extraction, our extractor sets this `SampleExtractionerror`. Fixes: grafana#6713 * Add it on changelog Signed-off-by: Kaviraj <[email protected]> * Fix linter Signed-off-by: Kaviraj <[email protected]> * Add more tests and fix the edge case with equality to empty strings Signed-off-by: Kaviraj <[email protected]> * typo on the comment Signed-off-by: Kaviraj <[email protected]> * handle this edge case during metric extraction. Not in label filter Signed-off-by: Kaviraj <[email protected]> * Fix linter Signed-off-by: Kaviraj <[email protected]> * Remove commented code Signed-off-by: Kaviraj <[email protected]> * Update changelog Signed-off-by: Kaviraj <[email protected]>

pull-request-size bot added the size/M label Jul 25, 2022

kavirajk commented Jul 25, 2022

View reviewed changes

kavirajk force-pushed the kavirajk/logql-label-check-bug branch from 7b0d82a to d0fbf3c Compare July 25, 2022 22:55

Add it on changelog

a83da68

Signed-off-by: Kaviraj <[email protected]>

pull-request-size bot added size/L and removed size/M labels Jul 25, 2022

Fix linter

8120941

Signed-off-by: Kaviraj <[email protected]>

kavirajk force-pushed the kavirajk/logql-label-check-bug branch from 70531ec to 8120941 Compare July 25, 2022 23:05

kavirajk marked this pull request as ready for review July 25, 2022 23:19

kavirajk requested a review from a team as a code owner July 25, 2022 23:19

kavirajk requested a review from owen-d July 25, 2022 23:19

ssncferreira reviewed Jul 26, 2022

View reviewed changes

pkg/logql/log/label_filter_test.go Show resolved Hide resolved

pkg/logql/log/metrics_extraction_test.go Show resolved Hide resolved

Add more tests and fix the edge case with equality to empty strings

bab6ca0

Signed-off-by: Kaviraj <[email protected]>

kavirajk requested a review from cyriltovena July 27, 2022 08:41

typo on the comment

a50c635

Signed-off-by: Kaviraj <[email protected]>

ssncferreira approved these changes Jul 27, 2022

View reviewed changes

cyriltovena reviewed Jul 28, 2022

View reviewed changes

kavirajk mentioned this pull request Jul 29, 2022

Issue using label filter prior to unwrapping the same field #6713

Closed

handle this edge case during metric extraction. Not in label filter

40c5953

Signed-off-by: Kaviraj <[email protected]>

kavirajk changed the title ~~fix(logql): Make StringLabelfilter work as expected~~ fix(logql): Make LabelSampleExtractor work as expected when label doesn't exist in logline Aug 1, 2022

kavirajk added 2 commits August 1, 2022 15:45

Fix linter

b915271

Signed-off-by: Kaviraj <[email protected]>

Remove commented code

b24112a

Signed-off-by: Kaviraj <[email protected]>

Merge branch 'main' into kavirajk/logql-label-check-bug

6568360

kavirajk requested a review from cyriltovena August 1, 2022 14:00

cyriltovena approved these changes Aug 3, 2022

View reviewed changes

Update changelog

2f9b01e

Signed-off-by: Kaviraj <[email protected]>

kavirajk merged commit 7b89f20 into main Aug 3, 2022

kavirajk deleted the kavirajk/logql-label-check-bug branch August 3, 2022 07:48

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

fix(logql): Make `LabelSampleExtractor` work as expected when label doesn't exist in logline #6766

fix(logql): Make `LabelSampleExtractor` work as expected when label doesn't exist in logline #6766

kavirajk commented Jul 25, 2022 •

edited

Loading

kavirajk Jul 25, 2022

grafanabot commented Jul 25, 2022

ssncferreira left a comment

grafanabot commented Jul 27, 2022

grafanabot commented Jul 27, 2022

ssncferreira left a comment

cyriltovena Jul 28, 2022 •

edited

Loading

kavirajk Jul 28, 2022 •

edited

Loading

owen-d Jul 28, 2022

kavirajk Aug 1, 2022

grafanabot commented Aug 1, 2022

grafanabot commented Aug 1, 2022

grafanabot commented Aug 1, 2022

cyriltovena left a comment

grafanabot commented Aug 3, 2022

fix(logql): Make LabelSampleExtractor work as expected when label doesn't exist in logline #6766

fix(logql): Make LabelSampleExtractor work as expected when label doesn't exist in logline #6766

Conversation

kavirajk commented Jul 25, 2022 • edited Loading

kavirajk Jul 25, 2022

Choose a reason for hiding this comment

grafanabot commented Jul 25, 2022

ssncferreira left a comment

Choose a reason for hiding this comment

grafanabot commented Jul 27, 2022

grafanabot commented Jul 27, 2022

ssncferreira left a comment

Choose a reason for hiding this comment

cyriltovena Jul 28, 2022 • edited Loading

Choose a reason for hiding this comment

kavirajk Jul 28, 2022 • edited Loading

Choose a reason for hiding this comment

owen-d Jul 28, 2022

Choose a reason for hiding this comment

kavirajk Aug 1, 2022

Choose a reason for hiding this comment

grafanabot commented Aug 1, 2022

grafanabot commented Aug 1, 2022

grafanabot commented Aug 1, 2022

cyriltovena left a comment

Choose a reason for hiding this comment

grafanabot commented Aug 3, 2022

fix(logql): Make `LabelSampleExtractor` work as expected when label doesn't exist in logline #6766

fix(logql): Make `LabelSampleExtractor` work as expected when label doesn't exist in logline #6766

kavirajk commented Jul 25, 2022 •

edited

Loading

cyriltovena Jul 28, 2022 •

edited

Loading

kavirajk Jul 28, 2022 •

edited

Loading