diff --git a/app/vmalert/manager.go b/app/vmalert/manager.go
index a15b4c97e0..b8ae851c9b 100644
--- a/app/vmalert/manager.go
+++ b/app/vmalert/manager.go
@@ -156,11 +156,14 @@ func (m *manager) update(ctx context.Context, groupsCfg []config.Group, restore
 		var wg sync.WaitGroup
 		for _, item := range toUpdate {
 			wg.Add(1)
+			// cancel evaluation so the Update will be applied as fast as possible.
+			// it is important to call InterruptEval before the update, because cancel fn
+			// can be re-assigned during the update.
+			item.old.InterruptEval()
 			go func(old *rule.Group, new *rule.Group) {
 				old.UpdateWith(new)
 				wg.Done()
 			}(item.old, item.new)
-			item.old.InterruptEval()
 		}
 		wg.Wait()
 	}
diff --git a/docs/CHANGELOG.md b/docs/CHANGELOG.md
index c7f75c6b3d..09f2dca29a 100644
--- a/docs/CHANGELOG.md
+++ b/docs/CHANGELOG.md
@@ -71,7 +71,8 @@ The sandbox cluster installation is running under the constant load generated by
 * BUGFIX: `vmstorage`: properly expire `storage/prefetchedMetricIDs` cache. Previously this cache was never expired, so it could grow big under [high churn rate](https://docs.victoriametrics.com/FAQ.html#what-is-high-churn-rate). This could result in increasing CPU load over time.
 * BUGFIX: [vmalert](https://docs.victoriametrics.com/vmalert.html): check `-external.url` schema when starting vmalert, must be `http` or `https`. Before, alertmanager could reject alert notifications if `-external.url` contained no or wrong schema.
 * BUGFIX: [vmalert](https://docs.victoriametrics.com/vmalert.html): automatically add `exported_` prefix for original evaluation result label if it's conflicted with external or reserved one, previously it was overridden. See [this issue](https://github.com/VictoriaMetrics/VictoriaMetrics/issues/5161).
-* BUGFIX: [vmalert](https://docs.victoriametrics.com/vmalert.html): autogenerate `ALERTS_FOR_STATE` time series for alerting rules with `for: 0`. Previously, `ALERTS_FOR_STATE` was generated only for alerts with `for > 0`. The change aligns with Prometheus behavior. See more details in [this issue](https://github.com/VictoriaMetrics/VictoriaMetrics/issues/5648). 
+* BUGFIX: [vmalert](https://docs.victoriametrics.com/vmalert.html): autogenerate `ALERTS_FOR_STATE` time series for alerting rules with `for: 0`. Previously, `ALERTS_FOR_STATE` was generated only for alerts with `for > 0`. The change aligns with Prometheus behavior. See more details in [this issue](https://github.com/VictoriaMetrics/VictoriaMetrics/issues/5648).
+* BUGFIX: [vmalert](https://docs.victoriametrics.com/vmalert.html): fix data race during hot-config reload. The result of the race could cause all rules from updated group to fail with `context cancelled` error.
 * BUGFIX: [MetricsQL](https://docs.victoriametrics.com/MetricsQL.html): consistently sort results for `q1 or q2` query, so they do not change colors with each refresh in Grafana. See [this issue](https://github.com/VictoriaMetrics/VictoriaMetrics/issues/5393).
 * BUGFIX: [MetricsQL](https://docs.victoriametrics.com/MetricsQL.html): properly return results from [bottomk](https://docs.victoriametrics.com/MetricsQL.html#bottomk) and `bottomk_*()` functions when some of these results contain NaN values. See [this issue](https://github.com/VictoriaMetrics/VictoriaMetrics/issues/5506). Thanks to @xiaozongyang for [the fix](https://github.com/VictoriaMetrics/VictoriaMetrics/pull/5509).
 * BUGFIX: [MetricsQL](https://docs.victoriametrics.com/MetricsQL.html): properly handle queries, which wrap [rollup functions](https://docs.victoriametrics.com/MetricsQL.html#rollup-functions) with multiple arguments without explicitly specified lookbehind window in square brackets into [aggregate functions](https://docs.victoriametrics.com/MetricsQL.html#aggregate-functions). For example, `sum(quantile_over_time(0.5, process_resident_memory_bytes))` was resulting to `expecting at least 2 args to ...; got 1 args` error. Thanks to @atykhyy for [the pull request](https://github.com/VictoriaMetrics/VictoriaMetrics/pull/5414).