deardeng commented on PR #67322:
URL: https://github.com/apache/doris/pull/67322#issuecomment-5537224875

   Hi @qzsee, could you add some background?
   
   1. Did this come from a real cluster issue, and what was the symptom?
   2. How did you find it — code reading, or logs/metrics?
   
   Reading the code, the repeated `registerWaterShedTxnId` looks more like 
wasted work than a correctness bug: the oldest entry carries the smallest 
watershed txn id, so `setActive(false)` still fires at the right time and the 
later duplicates are idempotent — decommission timing and outcome should be 
unchanged. It only really hurts when a decommission is stuck for a long time: N 
grows as (stuck_seconds / 10), and each `CloudUpgradeMgr` round then does N x 
`getAllTables()` plus N x `checkTxnConflict` RPCs to MS. Is that what you hit, 
or is there a worse failure mode I am missing?
   
   One note on the description: point 2 does not hold for the current code — 
since the guard checks `isDecommissioned()`, every round re-enters and retries 
anyway. Moving `setDecommissioning(true)` into the `try` is needed *because of* 
the guard change, not a pre-existing bug. Worth rewording so the two changes 
read as coupled.


-- 
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]


---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]

Reply via email to