DanielLeens commented on PR #12161:
URL: https://github.com/apache/seatunnel/pull/12161#issuecomment-5569391373

   +1 on this, @SEZ9 — good extension of what I flagged. Issue 2 is the one I 
missed: I focused on whether the per-member wait was bounded (it is, via 
`invocationFuture.get(3, SECONDS)`) but didn't trace that 
`members.stream().map(...)` (BaseService.java:1407-1420) dispatches 
`sendOperationToMemberNode` and blocks on that same member's future inside one 
lambda before moving to the next element — so the next member's RPC isn't even 
sent until the previous one returns or times out. I confirmed that against the 
current head (`88cc6b851818`): the send and the `get()` are both inside the 
same stream element. That means the endpoint is still worst-case `3s × member 
count`, which undercuts the PR title's "bound REST fanout" claim more directly 
than what I caught (I only flagged the missing test coverage and the 
non-propagating `cancel(true)`, not the sequencing itself). 
Dispatch-all-then-await-against-a-shared-deadline (or 
`CompletableFuture.allOf(...).get(3, SECONDS)`) is the right fi
 x.
   
   Issue 4 is also a solid catch and compounds Issue 2: once one member throws 
`InterruptedException`, the flag stays set for the rest of the still-sequential 
stream, so every remaining member fails instantly and silently with nothing 
logged — worth calling out as its own item alongside the High-severity 
test-coverage gap I raised.
   
   Combining our two passes, I'd treat your Issues 1, 2 and 4 together with my 
Issue 3 (zero test coverage on the timeout/interrupt/cancel branch) as the 
blocking set for this round; the Spotless/doc/locale-test items (your 3, 5, 6, 
my 1) are straightforward mechanical fixes. Thanks for the thorough pass — this 
needs another revision before merge.


-- 
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]

Reply via email to