Update vendored mimir-prometheus #8305
Conversation
| Offset: selector.Offset, | ||
| OriginalOffset: selector.OriginalOffset, | ||
| Timestamp: copyTimestamp(selector.Timestamp), | ||
| SkipHistogramBuckets: selector.SkipHistogramBuckets, |
There was a problem hiding this comment.
Note to reviewers, in particular @krajorama: shardVectorSelector() is meant to copy the input VectorSelector, only modifying the LabelMatchers. For this reason, I think it's safer (and more consistent) to also copy SkipHistogramBuckets.
There was a problem hiding this comment.
From how the skip histogram bucket works, you cannot really break things by omitting it. Omitting it would just mean that we copy buckets as well in the histogram iterator - so more data.
Conversely, adding it would only break if in the sharded query we started to look at buckets suddenly, which would mean adding a function call like histogram_quantile, histogram_fraction, histogram_stddev, histogram_stdvar. But we never do, we just modify the query to add sum(), which is safe.
There was a problem hiding this comment.
But this question does point out that we need a generic way to handle optimizations on the AST so we can just re-run the optimization without having to worry about this on a case by case basis @beorn7
There was a problem hiding this comment.
No better insights to share from my side, unfortunately. It sounds to me we have to check optimization on a case by case basis to see how they interact with sharded query evaluation.
pracucci
force-pushed
the
update-mimir-prometheus
branch
from
7655691
to
000b35b
Compare
|
I think I've found an issue in |
| @@ -36,6 +36,7 @@ const ( | |||
| "http_config": { | |||
| "enable_http2": true, | |||
| "follow_redirects": true, | |||
| "http_headers": null, | |||
There was a problem hiding this comment.
Note to reviewers: this is the default value. Added so that the check done by TestMultitenantAlertmanager_GetUserGrafanaConfig pass.
Signed-off-by: Marco Pracucci <marco@pracucci.com>
Signed-off-by: Marco Pracucci <marco@pracucci.com>
Signed-off-by: Marco Pracucci <marco@pracucci.com>
Signed-off-by: Marco Pracucci <marco@pracucci.com>
Signed-off-by: Marco Pracucci <marco@pracucci.com>
Signed-off-by: Marco Pracucci <marco@pracucci.com>
Signed-off-by: Marco Pracucci <marco@pracucci.com>
Signed-off-by: Marco Pracucci <marco@pracucci.com>
Signed-off-by: Marco Pracucci <marco@pracucci.com>
pracucci
force-pushed
the
update-mimir-prometheus
branch
from
f6c94fc
to
a487bbc
Compare
| github.com/dennwc/varint v1.0.0 | ||
| github.com/golang/groupcache v0.0.0-20210331224755-41bb18bfe9da | ||
| github.com/google/go-cmp v0.6.0 | ||
| github.com/google/go-github/v57 v57.0.0 | ||
| github.com/google/uuid v1.6.0 | ||
| github.com/grafana-tools/sdk v0.0.0-20220919052116-6562121319fc | ||
| github.com/grafana/alerting v0.0.0-20240605124151-5d695b88086a | ||
| github.com/grafana/regexp v0.0.0-20221122212121-6b5c0a4cb7fd | ||
| github.com/grafana/regexp v0.0.0-20240518133315-a468a5bfb3bc |
There was a problem hiding this comment.
No effect due to replace directive with v0.0.0-20240531075221-3685f1377d7b
There was a problem hiding this comment.
looks ok. what we're using is coming from newer optimization branch by @bboreham
| github.com/hashicorp/golang-lru/v2 v2.0.7 | ||
| github.com/hashicorp/vault/api v1.10.0 | ||
| github.com/mitchellh/colorstring v0.0.0-20190213212951-d06e56a500db | ||
| github.com/munnerz/goautoneg v0.0.0-20191010083416-a7dc8b61c822 | ||
| github.com/prometheus/procfs v0.12.0 | ||
| github.com/prometheus/procfs v0.15.1 |
There was a problem hiding this comment.
This is a big jump and the library does not guarantee backwards compatibility. That said, their release notes do not call out any breaking change.
What this PR does
Updating vendored mimir-prometheus, getting the changes synched in grafana/mimir-prometheus#641 (see list of changes there).
Notes:
Paramsremoval while I kept the other optimizationgrpc.Dial()andgrpc.DialContext()have been deprecated, but they will be supported in gRPC 1.x. We can defer handling it once we'll be close to migrate to gRPC 2.While working on this PR I've found a couple of issue, whose fixes have already merged upstream:
Which issue(s) this PR fixes or relates to
N/A
Checklist
CHANGELOG.mdupdated - the order of entries should be[CHANGE],[FEATURE],[ENHANCEMENT],[BUGFIX].about-versioning.mdupdated with experimental features.