Claus Ibsen created CAMEL-24196:
-----------------------------------
Summary: camel-aws2-sqs - QueueDoesNotExist swallowed, thread pool
leak, and batch chunking
Key: CAMEL-24196
URL: https://issues.apache.org/jira/browse/CAMEL-24196
Project: Camel
Issue Type: Bug
Components: camel-aws
Reporter: Claus Ibsen
Split from CAMEL-24157 (medium-severity findings, group 3).
Three remaining SQS findings that require careful design work:
*1. QueueDoesNotExistException silently swallowed when maxMessagesPerPoll > 10*
{{Sqs2Consumer:597,655-671}} — When {{maxMessagesPerPoll > 10}}, parallel
polling sends multiple requests. The first thread to catch
{{QueueDoesNotExistException}} calls {{firePollingError}} (registering 1
error), but remaining threads see {{isQueueMissing()}} and return empty without
firing errors. Error propagation at line 597 requires {{errorCount() ==
numberOfRequestsPerPoll}}, which never matches (1 ≠ N). The missing-queue error
is silently swallowed and the consumer returns empty forever. Works correctly
when {{maxMessagesPerPoll <= 10}}.
*2. Consumer stop/start cycle leaks PollingTask thread pool*
{{Sqs2Consumer:302-363,578-587}} — {{doStart()}} unconditionally creates a new
{{PollingTask}} with a new {{ExecutorService}} thread pool. Cleanup only
happens in {{doShutdown()}}, not {{doStop()}}. Each route stop/start cycle
leaks the previous thread pool.
*3. sendBatchMessage lacks 10-entry chunking*
{{Sqs2Producer:119-174}} — All entries collected into a single
{{sendMessageBatch()}} call. SQS hard-limits batches to 10 entries; 11+ throws
{{TooManyEntriesInBatchRequestException}}. Empty iterables and non-String
elements are also unhandled.
*Fix approach:*
- #1: Propagate error when any parallel request gets QueueDoesNotExist, not
only when all fail
- #2: Add {{doStop()}} override to shut down the PollingTask thread pool and
cancel the visibility extender
- #3: Chunk entries into groups of 10 before calling {{sendMessageBatch()}};
validate inputs
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
