Can I use same objstore on both metrics and alerts?

[kamontat]

Can I reused same objstore (specifically S3 bucket) for both metric and alert?

[dosubot[bot]]

Hey @kamontat! I'm here to help you with any bugs, questions, or contributions you have. Let's solve this together!

Using the same S3 bucket for both metrics and alerts in Thanos is possible, but there are several potential risks and conflicts to be aware of:

1. Concurrency Issues: The thanos compact command is not concurrency safe and should be deployed as a singleton against a bucket. Running multiple compactors against the same bucket can lead to overlap issues that need manual resolution.

2. Overlap Issues: Blocks with the same external labels that overlap in time can cause the compactor to halt, which is problematic if multiple compactors are used on the same bucket.

3. Vertical Compaction Risks: Configuration errors, such as uploading blocks with the same external labels from different Prometheis, can cause metrics to overlap and merge, rendering the series useless.

4. Data Deletion Risks: If retention settings are not properly configured, data might be deleted before downsampling is completed.

5. Eventual Consistency Challenges: Operating in an eventually consistent object storage environment can lead to issues like partial uploads. Thanos assumes a block is ready to be read only after a certain delay to handle eventual consistency.

6. Block Immutability and Deletion: Blocks are never edited once uploaded, and deletion operations are non-atomic, which can lead to temporary inconsistencies ^[1][2].

These considerations highlight the importance of careful configuration and management when using a shared object store for both metrics and alerts.

To continue talking to Dosu, mention @dosu.

---

^{Help Dosu learn! Give it feedback: Great Response | Irrelevant Answer | Incorrect Sources | Too Verbose | Hallucination | Other}

[kamontat]

@dosu Does thanos compact run ruler as well?

[dosubot[bot]]

No, Thanos Compactor and Thanos Ruler are separate components with distinct functions. The Thanos Compactor is responsible for managing the lifecycle of block data in object storage, including compaction, downsampling, and retention ^[1]. On the other hand, the Thanos Ruler is a rule evaluation engine that evaluates Prometheus alerting and recording rules and manages alert notifications ^[2]. They do not run each other or have direct dependencies in their operations.

To continue talking to Dosu, mention @dosu.

---

^{Help Dosu learn! Give it feedback: Great Response | Irrelevant Answer | Incorrect Sources | Too Verbose | Hallucination | Other}

[kamontat]

@dosu If I write Thanos Ruler data to same store as Thanos metrics, will it works?

[dosubot[bot]]

Yes, writing Thanos Ruler data to the same object store as Thanos metrics will work. Thanos Ruler stores its data using the Prometheus 2.0 storage format, which is compatible with the data format used by Thanos metrics ^[1]. However, be mindful of potential implications such as data retention and compaction. Ensure that only one instance of the Thanos Compactor runs against the bucket to avoid concurrency issues, and configure retention settings appropriately to manage data lifecycle ^[2].

To continue talking to Dosu, mention @dosu.

---

^{Help Dosu learn! Give it feedback: Great Response | Irrelevant Answer | Incorrect Sources | Too Verbose | Hallucination | Other}