pbacsko commented on PR #1041: URL: https://github.com/apache/yunikorn-k8shim/pull/1041#issuecomment-4734174444
Is there a corresponding change for the core? The `bind.failure.reason` is not consumed by the core at all. Am I missing something? This was my initial impression. Then I admit I got lazy and asked Opus and GPT-5.5 to check this patch and both concluded that this change alone is NOT enough. There is no code path in the core which would make the transition from "allocated" -> "pending". Here is the relevant parts from Opus: ``` There is no branch handling the allocated → unallocated transition. The core keeps the allocation on the failed node, never calls DeallocateAsk, never frees the node/queue resources, and never re-allocates. The two existing DeallocateAsk call sites in the core are only for placeholder-replacement reversal on node removal — unrelated to this path. Additionally, the BindFailureReasonTag the shim attaches is never read anywhere in yunikorn-core (no references to bind.failure/BindFailure exist there), so it is silently dropped. ``` GPT-5.5 reached the same conclusion. They ran in high/extra high mode. -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected]
