security icon indicating copy to clipboard operation
security copied to clipboard

[BUG/Discussion] Health endpoint does not provide real information about final intialization state of the cluster.

Open willyborankin opened this issue 1 year ago • 2 comments

What is the bug? The health REST endpoint does not provide the real information about security initialization status. In the current logic, it returns only health for the predefined or random coordinator node, but not for the whole cluster.

What is the expected behavior? One of the solutions is to add a new transport action to check the status on each node and, based on the responses from nodes, calculate the final status UP/DOWN

Another solution is to move cluster to the global blocking state and release the global block as soon as all nodes initialized the security config.

willyborankin avatar May 15 '24 20:05 willyborankin

[Triage] @willyborankin thank you for filing this issue! It seems like this is a bug that can be fixed by calling a transport API that would only work if security had been initialized in all nodes - this bug could also have been related to several other bugs that have come up in the past regarding race conditions during cluster startup.

derek-ho avatar May 20 '24 15:05 derek-ho

How do other plugins/components that use system index for storing the states deal with this?

varun-lodaya avatar Jun 03 '24 07:06 varun-lodaya