[
https://issues.apache.org/jira/browse/CAMEL-24193?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Work on CAMEL-24193 started by Andrea Cosentino.
------------------------------------------------
> camel-aws-secrets-manager: SQS-notification secret-refresh NPEs on missing
> queue and never drains the queue
> -----------------------------------------------------------------------------------------------------------
>
> Key: CAMEL-24193
> URL: https://issues.apache.org/jira/browse/CAMEL-24193
> Project: Camel
> Issue Type: Bug
> Components: camel-aws
> Reporter: Andrea Cosentino
> Assignee: Andrea Cosentino
> Priority: Major
> Fix For: 4.14.9, 4.18.4, 4.22.0
>
>
> In the SQS-notification branch of CloudTrailReloadTriggerTask (the
> aws-secret-refresh task with useSqsNotification=true):
> 1. *NPE on a receive failure.* {{QueueDoesNotExistException}} from
> {{sqsClient.receiveMessage(...)}} is caught and only logged, leaving
> {{messageResult}} null, which is then dereferenced by {{for (Message message
> : messageResult.messages())}} — every poll throws NPE (swallowed by the outer
> catch and logged as an error). Any transient receive failure that leaves
> messageResult null has the same effect. Code ~lines 324-332.
> 2. *Queue is never drained.* {{deleteMessage}} runs only inside {{if
> (matchSecret) { if (isReloadEnabled()) { ... } }}} and the loop {{break}}s on
> the first match. Non-matching messages, and all messages when
> reloadEnabled=false, are never deleted and are redelivered on every poll
> (unbounded queue growth / cost); only the first matching message in a batch
> is handled. Code ~lines 347-360.
> 3. *ObjectMapper allocated per message* inside the loop (~line 333) instead
> of reused.
> Fix direction: guard against a null receive result; parse with a single
> reused ObjectMapper; process every message in the batch and delete each one
> after examining it (matched -> record update / trigger reload; unmatched or
> malformed -> discard) so the queue is drained. A malformed message body
> should be logged and discarded rather than aborting the whole batch.
> Not changing here (noted for follow-up): the CloudTrail-polling branch
> advances its cursor by newest-event-time + 1000ms (can skip same-second
> events), and matchSecret uses substring contains() (over-broad) — both are
> behavior-sensitive and out of scope.
> Present on camel-4.18.x and camel-4.14.x; backport to both.
> Found during an agent-assisted audit of the AWS components; verified against
> the cited code.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
