-
Notifications
You must be signed in to change notification settings - Fork 1.1k
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Upgrading v1.63.0 -> v1.65.0 leads to unexpeced huge datapoint #1628
Comments
And a question: how does vmselect selects vmstorage to request? From my reading of the code, vmselct will read all vmstorage until it gets what it wants ? The reason I ask this is when I request directly from vmselct it gives persistent results through the update buttion at the right corner of grafana despite all vmselect share the same cmdline config of storageNodes, which should give different results, since it requests from all vmstorage at once? |
@sequix , thanks for the detailed analysis of the issue! VictoriaMetrics components emit and process e.g. Prometheus staleness marks starting from v1.64.0 - see https://docs.victoriametrics.com/CHANGELOG.html#v1640 and #1526 . Previous versions of VictoriaMetrics (v1.63.0 or lower) just drop the staleness marks during data ingestion (at So cluster components of VictoriaMetrics should be upgraded in the following order:
Correct! |
Clears pretty much my confusions, but it does not explain this test result if
|
seems so, close it. |
Describe the bug
I was upgrading our VictoriaMetrics in the follwing order:
After upgrading 1 machine, I got the following results at some random timeserie and random vmselct endpoint (which cause pretty much alerts sent...):
I noticed the vmagent went down at the same instant:
After upgrading all machines, the big datapoint disappeared:
To Reproduce
Use the following program to request different clusetr version setup.
The cmdline for every setup:
And get the results with:
Results:
Expected behavior
I do not exactly know what is expected for the low version components, but alerts sent by these huge datapoints shoud be avoided.
Maybe an intrstruction on upgrading order & pitfalls,or heads up at the release info is needed to articulate these behavior and how to avoid them.
The text was updated successfully, but these errors were encountered: