REPLICATOR: Add TimePartitionMetaFilter to thanos replicate. (thanos-io#2979)

* Add TimePartitionMetaFilter to thanos replicate.

This will allow time based replication.

Signed-off-by: Kevin Hellemun <[email protected]>

* Update docs and changelog.

Signed-off-by: Kevin Hellemun <[email protected]>

* Fix lint

Signed-off-by: Kevin Hellemun <[email protected]>

* Add posibility to delete blocks older than max time

Signed-off-by: Kevin Hellemun <[email protected]>

* Fix metric naming.

Signed-off-by: Kevin Hellemun <[email protected]>

* Fix order of arguments for max and min time.

Signed-off-by: Kevin Hellemun <[email protected]>

* Remove deletion of blocks from replicator.

Signed-off-by: Kevin Hellemun <[email protected]>

* Fix import formatting

Signed-off-by: Kevin Hellemun <[email protected]>

* Revert unneeded changes due to replicator no longer deleting blocs

Signed-off-by: Kevin Hellemun <[email protected]>

* Fix typo

Signed-off-by: Kevin Hellemun <[email protected]>
Signed-off-by: Oghenebrume50 <[email protected]>

Author: OGKevin

CHANGELOG.md

@@ -22,6 +22,7 @@ We use *breaking :warning:* to mark changes that are not backward compatible (re
 - [#3312](https://github.com/thanos-io/thanos/pull/3312) s3: add list_objects_version config option for compatibility.
 - [#3356](https://github.com/thanos-io/thanos/pull/3356) Query Frontend: Add a flag to disable step alignment middleware for query range.
 - [#3378](https://github.com/thanos-io/thanos/pull/3378) Ruler: added the ability to send queries via the HTTP method POST. Helps when alerting/recording rules are extra long because it encodes the actual parameters inside of the body instead of the URI. Thanos Ruler now uses POST by default unless `--query.http-method` is set `GET`.
+- [#2979](https://github.com/thanos-io/thanos/pull/2979) Replicator: Add the ability to replicate blocks within a time frame by passing --min-time and --max-time
 
 ### Fixed
 - [#3257](https://github.com/thanos-io/thanos/pull/3257) Ruler: Prevent Ruler from crashing when using default DNS to lookup hosts that results in "No such hosts" errors.

cmd/thanos/tools_bucket.go

@@ -36,6 +36,7 @@ import (
 	"github.com/thanos-io/thanos/pkg/extprom"
 	extpromhttp "github.com/thanos-io/thanos/pkg/extprom/http"
 	"github.com/thanos-io/thanos/pkg/logging"
+	"github.com/thanos-io/thanos/pkg/model"
 	"github.com/thanos-io/thanos/pkg/objstore"
 	"github.com/thanos-io/thanos/pkg/objstore/client"
 	"github.com/thanos-io/thanos/pkg/prober"
@@ -444,6 +445,10 @@ func registerBucketReplicate(app extkingpin.AppClause, objStoreConfig *extflag.P
 	compactions := cmd.Flag("compaction", "Only blocks with these compaction levels will be replicated. Repeated flag.").Default("1", "2", "3", "4").Ints()
 	matcherStrs := cmd.Flag("matcher", "Only blocks whose external labels exactly match this matcher will be replicated.").PlaceHolder("key=\"value\"").Strings()
 	singleRun := cmd.Flag("single-run", "Run replication only one time, then exit.").Default("false").Bool()
+	minTime := model.TimeOrDuration(cmd.Flag("min-time", "Start of time range limit to replicate. Thanos Replicate will replicate only metrics, which happened later than this value. Option can be a constant time in RFC3339 format or time duration relative to current time, such as -1d or 2h45m. Valid duration units are ms, s, m, h, d, w, y.").
+		Default("0000-01-01T00:00:00Z"))
+	maxTime := model.TimeOrDuration(cmd.Flag("max-time", "End of time range limit to replicate. Thanos Replicate will replicate only metrics, which happened earlier than this value. Option can be a constant time in RFC3339 format or time duration relative to current time, such as -1d or 2h45m. Valid duration units are ms, s, m, h, d, w, y.").
+		Default("9999-12-31T23:59:59Z"))
 
 	cmd.Setup(func(g *run.Group, logger log.Logger, reg *prometheus.Registry, tracer opentracing.Tracer, _ <-chan struct{}, _ bool) error {
 		matchers, err := replicate.ParseFlagMatchers(*matcherStrs)
@@ -469,6 +474,8 @@ func registerBucketReplicate(app extkingpin.AppClause, objStoreConfig *extflag.P
 			objStoreConfig,
 			toObjStoreConfig,
 			*singleRun,
+			minTime,
+			maxTime,
 		)
 	})
 }

docs/components/tools.md

@@ -472,6 +472,22 @@ Flags:
       --matcher=key="value" ...  Only blocks whose external labels exactly match
                                  this matcher will be replicated.
       --single-run               Run replication only one time, then exit.
+      --min-time=0000-01-01T00:00:00Z
+                                 Start of time range limit to replicate. Thanos
+                                 Replicate will replicate only metrics, which
+                                 happened later than this value. Option can be a
+                                 constant time in RFC3339 format or time
+                                 duration relative to current time, such as -1d
+                                 or 2h45m. Valid duration units are ms, s, m, h,
+                                 d, w, y.
+      --max-time=9999-12-31T23:59:59Z
+                                 End of time range limit to replicate. Thanos
+                                 Replicate will replicate only metrics, which
+                                 happened earlier than this value. Option can be
+                                 a constant time in RFC3339 format or time
+                                 duration relative to current time, such as -1d
+                                 or 2h45m. Valid duration units are ms, s, m, h,
+                                 d, w, y.
 
 ```
 

pkg/replicate/replicator.go

@@ -10,6 +10,8 @@ import (
 	"strings"
 	"time"
 
+	thanosmodel "github.com/thanos-io/thanos/pkg/model"
+
 	"github.com/go-kit/kit/log"
 	"github.com/go-kit/kit/log/level"
 	"github.com/oklog/run"
@@ -80,6 +82,7 @@ func RunReplicate(
 	fromObjStoreConfig *extflag.PathOrContent,
 	toObjStoreConfig *extflag.PathOrContent,
 	singleRun bool,
+	minTime, maxTime *thanosmodel.TimeOrDurationValue,
 ) error {
 	logger = log.With(logger, "component", "replicate")
 
@@ -161,7 +164,15 @@ func RunReplicate(
 	replicationRunDuration.WithLabelValues(labelSuccess)
 	replicationRunDuration.WithLabelValues(labelError)
 
-	fetcher, err := thanosblock.NewMetaFetcher(logger, 32, fromBkt, "", reg, nil, nil)
+	fetcher, err := thanosblock.NewMetaFetcher(
+		logger,
+		32,
+		fromBkt,
+		"",
+		reg,
+		[]thanosblock.MetadataFilter{thanosblock.NewTimePartitionMetaFilter(*minTime, *maxTime)},
+		nil,
+	)
 	if err != nil {
 		return errors.Wrapf(err, "create meta fetcher with bucket %v", fromBkt)
 	}

pkg/replicate/scheme.go

@@ -241,7 +241,7 @@ func (rs *replicationScheme) ensureBlockIsReplicated(ctx context.Context, id uli
 			// If the origin meta file content and target meta file content is
 			// equal, we know we have already successfully replicated
 			// previously.
-			level.Debug(rs.logger).Log("msg", "skipping block as already replicated", "block_uuid", id.String())
+			level.Debug(rs.logger).Log("msg", "skipping block as already replicated", "block_uuid", blockID)
 			rs.metrics.blocksAlreadyReplicated.Inc()
 
 			return nil

pkg/replicate/scheme_test.go

@@ -315,15 +315,7 @@ func TestReplicationSchemeAll(t *testing.T) {
 		fetcher, err := block.NewMetaFetcher(logger, 32, objstore.WithNoopInstr(originBucket), "", nil, nil, nil)
 		testutil.Ok(t, err)
 
-		r := newReplicationScheme(
-			logger,
-			newReplicationMetrics(nil),
-			filter,
-			fetcher,
-			objstore.WithNoopInstr(originBucket),
-			targetBucket,
-			nil,
-		)
+		r := newReplicationScheme(logger, newReplicationMetrics(nil), filter, fetcher, objstore.WithNoopInstr(originBucket), targetBucket, nil)
 
 		err = r.execute(ctx)
 		testutil.Ok(t, err)