Salvatore Mongiardo created CAMEL-26779:
-------------------------------------------
Summary: camel-snmp - the consumer can run its first poll before
it is started, and fails with a NullPointerException
Key: CAMEL-26779
URL: https://issues.apache.org/jira/browse/CAMEL-26779
Project: Camel
Issue Type: Bug
Components: camel-snmp
Affects Versions: 4.22.1
Reporter: Salvatore Mongiardo
Fix For: 4.24.0, 4.22.3
SnmpOIDPoller.doStart() calls super.doStart() first.
ScheduledPollConsumer.doStart() schedules the poll task, and only afterwards
doStart() creates the Snmp instance, the USM, the target and (in 4.22.x) the
request PDU, and starts the transport listener.
If this initialization is slow, the first scheduled poll runs in between and
finds the fields not yet set. In 4.22.x the poll fails with:
{noformat}
WARN SnmpOIDPoller : Failed polling endpoint: snmp://udp:<host>/<port>. Will
try again at next poll.
Caused by: [java.lang.NullPointerException - Cannot invoke
"org.snmp4j.PDU.clear()" because "this.pdu" is null]
at
org.apache.camel.component.snmp.SnmpOIDPoller.poll(SnmpOIDPoller.java:103)
{noformat}
Observed with Camel 4.22.1 on Windows x86_64 and JDK 21, where the gap between
the route start and "Started OID poller" was about 17 seconds. It happened in 3
of 3 runs.
On main, poll() creates a new PDU for each poll (CAMEL-25924), so the PDU is no
longer null. The race remains for snmp and target, which are still created
after super.doStart() has scheduled the poll.
The consumer recovers on the next poll cycle, but the first poll fails with a
WARN and the first expected message is delayed or lost.
Proposed change: in SnmpOIDPoller.doStart(), initialize snmp, the USM, the
target and the transport first, call transport.listen(), and call
super.doStart() last. The scheduler then starts only when the consumer is ready
to poll.
Fix for main and for camel-4.22.x.
--
This message was sent by Atlassian Jira
(v10.3.26#10030026)
