[Bug] cloudberry cluster becoming unavailable

[yudhvir12afiniti]

Apache Cloudberry version

apache cloudberry 2.0.0

What happened

Hi Team,

First there was segment crash, where the mirror became active and primary is in failed state, then on top of that there is another crash on the same segment where primary became active and mirror went into failed state.

As soon as the primary became active and mirror went to failed state for the second time crash, the cloudberry cluster became unavailable because the Standby. Signal file was not removed from the primary segment instance data directory and we need to stop and start the cloudberry cluster to make it available.

Can you please kindly provide your input\solution on this.

What you think should happen instead

No response

How to reproduce

first crash a segment and then immediately conduct one more crash on the same segment,

Operating System

OEL 9.7

Anything else

No response

Are you willing to submit PR?

• Yes, I am willing to submit a PR!

Code of Conduct

• I agree to follow this project's Code of Conduct.

[github-actions]

Hi, @yudhvir12afiniti welcome!🎊 Thanks for taking the time to point this out.🙌

[liang8283]

Hi @yudhvir12afiniti ,

Did you run the gprecoverseg to recover the failed primary after the first time crash? If not, this is expected behavior the cluster becomes unavailable after the second time crash on the same segment because both primary and mirror goes down in this case. Either primary or mirror has to be up and running for one segment. So you need to run gprecoverseg to recover the failed segment after the first time crash to avoid such kind of issue.

[yudhvir12afiniti]

Thank you @liang8283, I ddin't ran the gprecoverseg to recover the failed primary after the first time crash. But some times this behavior is not the case even though second time crash on the same segment.