OPERATIONAL DEFECT DATABASE
...
...
According to the design doc for causal consistency a non-sharded replica set should still honor afterClusterTime even in the absence of $clusterTime gossiping. But in read_concern.cpp there is a check that the logical clock has advanced to afterClusterTime which is performed before waiting for that cluster time to replicate through the oplog. The result is that readConcern.afterClusterTime fails consistently on secondary members of replica sets that are not part of a sharded cluster, but succeeds when part of a sharded cluster.
jeff.yemin commented on Thu, 31 Aug 2017 17:44:00 +0000: Here's logging of one failure case: Server: localhost:27017 Start: 08:18:39.500 Command: { "insert" : "causal", "ordered" : true, "documents" : [{ "_id" : 580 }], "$db" : "test", "lsid" : { "id" : { "$binary" : "UJp6HUyiSEWnCEz2XAtyxA==", "$type" : "04" } } } End: 08:18:39.500 Response: { "n" : 1, "opTime" : { "ts" : { "$timestamp" : { "t" : 1504181919, "i" : 9 } }, "t" : { "$numberLong" : "66" } }, "electionId" : { "$oid" : "7fffffff0000000000000042" }, "ok" : 1.0, "operationTime" : { "$timestamp" : { "t" : 1504181919, "i" : 9 } } } Server: localhost:27018 Start: 08:18:39.553 Command: { "find" : "causal", "filter" : { "_id" : 580 }, "limit" : 1, "singleBatch" : true, "readConcern" : { "level" : "local", "afterClusterTime" : { "$timestamp" : { "t" : 1504181919, "i" : 9 } } }, "$db" : "test", "lsid" : { "id" : { "$binary" : "UJp6HUyiSEWnCEz2XAtyxA==", "$type" : "04" } }, "$readPreference" : { "mode" : "secondary" } } Failure: Command failed with error 72: 'readConcern afterClusterTime must not be greater than clusterTime value' on server jeff.fios-router.home:27018. The full response is { "ok" : 0.0, "errmsg" : "readConcern afterClusterTime must not be greater than clusterTime value", "code" : 72, "codeName" : "InvalidOptions" }
1. Create a standard replica set 2. Insert a document and record the operationTime from the response 3. Attempt to find that document on a secondary with a readConcern containing afterClusterTime whose value is the operationTime from the insert response Expected results: the find command succeeds and returns the requested document Actual results: the secondary returns this error: readConcern afterClusterTime must not be greater than clusterTime value", "code" : 72, "codeName" : "InvalidOptions"
Click on a version to see all relevant bugs
MongoDB Integration
Learn more about where this data comes from
Bug Scrub Advisor
Streamline upgrades with automated vendor bug scrubs
BugZero Enterprise
Wish you caught this bug sooner? Get proactive today.
