shashank created CAMEL-26992:
--------------------------------

             Summary: camel-mongodb - a change stream that the consumer opens 
again before any exchange completed starts at the current time and skips the 
changes made in between
                 Key: CAMEL-26992
                 URL: https://issues.apache.org/jira/browse/CAMEL-26992
             Project: Camel
          Issue Type: Bug
          Components: camel-mongodb
            Reporter: shashank
            Assignee: shashank


Found by the MongoDB change streams DST world of CAMEL-26973 (branch 
{{camel-mongodb-dst-world}}), which leaves this case out of its default set: 
with {{-Ddst.mongodb.strictFirstCursor=true}}, 50 of 2000 seeds (first failing 
seed 12) end with changes of the collection that the route never got. 
Reproduced outside the DST harness with a plain unit test (below).

h3. What happens

When the change streams consumer has no resume token (no {{changeStreamToken}}, 
nothing in the {{changeStreamTokenRepository}}, no resume strategy), it opens 
its change stream without a resume option, so the stream starts at the current 
time. That is fine for the first cursor of a route start. But 
{{MongoDbChangeStreamsThread}} also opens a new cursor itself when the change 
stream fails, for example when a getMore fails and the resume of the driver 
fails too because the server cannot be reached for longer than the server 
selection timeout (the retry loop of CAMEL-25679). If no exchange completed 
yet, the consumer still has no resume token, so that new cursor also starts at 
the current time, and every change made since the first cursor was opened is 
skipped. Once an exchange completed, the consumer resumes after its token as 
expected.

h3. Expected

The consumer thread is still running, so a cursor it opens again should go on 
where the previous one stopped, as it does once there is a token. The MongoDB 
manual (Change Streams, "Resume a Change Stream") documents 
{{startAtOperationTime}}: "You can specify a startAtOperationTime to open the 
cursor at a particular point in time. If the specified starting point is in the 
past, it must be in the time range of the oplog." The driver does this itself 
for its own resume (change streams specification: the operationTime of the 
initial aggregate is saved and used as {{startAtOperationTime}} when there is 
no cached token), but the Java driver does not expose that time, and 
{{MongoChangeStreamCursor.getResumeToken()}} is only set once a getMore 
succeeded, which is exactly what fails here.

h3. Reproduction

New {{MongoDbChangeStreamsReopenTest}} (a mocked client whose change streams 
answer from a list of changes): one change before the route starts, then the 
route starts with no token; a and b are inserted, the first getMore fails 
({{MongoSocketReadException}}, the driver could not resume), and c is inserted 
after the consumer opened a new cursor. On main the route gets {{[c]}}: a and b 
are lost.

h3. Fix

Before it opens a change stream at the current time, the consumer sends a 
{{ping}} (with the read preference of the collection, so it goes where the 
change stream reads from) and keeps the {{operationTime}} of the reply (for 
read concern local, "the timestamp of the most recent entry in the oplog", 
{{db.runCommand}} reference), plus one increment. Until an exchange completed, 
a cursor that the consumer thread opens again starts with 
{{startAtOperationTime}} at that time, so it delivers the changes made since. 
The time is not stored: a route started again without a token still starts at 
the current time, as documented. If the oplog no longer goes back that far 
(ChangeStreamHistoryLost 286, or ChangeStreamFatalError 280 or 40576 on earlier 
4.2 releases) the consumer logs a WARN and starts at the current time as 
before. A server whose reply has no {{operationTime}} (a standalone server, 
which has no change streams) or that answers the ping with an error keeps the 
current behaviour. Branch 
{{camel-mongodb-change-streams-reopen-before-first-exchange-v2}}.

The DST world needs its fake to answer the {{ping}} and 
{{startAtOperationTime}} once both are in; with that and the fix, 
{{-Ddst.mongodb.strictFirstCursor=true}} passes 2000 of 2000 seeds and the 
default set 5000 of 5000.

_Filed with Claude Code on behalf of allthingssecurity._



--
This message was sent by Atlassian Jira
(v10.3.26#10030026)

Reply via email to