diff options
| author | Suganthi Mani <suganthi.mani@mongodb.com> | 2018-12-04 15:47:03 -0500 |
|---|---|---|
| committer | Suganthi Mani <suganthi.mani@mongodb.com> | 2019-01-23 11:08:09 -0500 |
| commit | 1a24b7d77d1271de19b61705b3890953d485978e (patch) | |
| tree | 99c74e9e89704d42f48c0279d4e042c6ea4cdbe0 | |
| parent | 5c7c6729c37514760fd34da462b6961a2e385417 (diff) | |
| download | mongo-1a24b7d77d1271de19b61705b3890953d485978e.tar.gz | |
SERVER-37910 Added a new serverStatus meteric "repl.apply.batchSize" to track oplog application batch size.
(cherry picked from commit cc102992f6fe9dd499edc9242b8a5be961ed35dd)
| -rw-r--r-- | jstests/replsets/server_status_metrics.js | 1 | ||||
| -rw-r--r-- | src/mongo/db/repl/sync_tail.cpp | 12 |
2 files changed, 12 insertions, 1 deletions
diff --git a/jstests/replsets/server_status_metrics.js b/jstests/replsets/server_status_metrics.js index 9b805b92e4c..59054f2172a 100644 --- a/jstests/replsets/server_status_metrics.js +++ b/jstests/replsets/server_status_metrics.js @@ -25,6 +25,7 @@ function testSecondaryMetrics(secondary, opCount, offset) { assert(ss.metrics.repl.preload.docs.num >= 0, "preload.indexes num missing"); assert(ss.metrics.repl.preload.indexes.totalMillis >= 0, "preload.indexes time missing"); + assert.eq(ss.metrics.repl.apply.batchSize, opCount + offset, "apply ops batch size mismatch"); assert(ss.metrics.repl.apply.batches.num > 0, "no batches"); assert(ss.metrics.repl.apply.batches.totalMillis >= 0, "missing batch time"); assert.eq(ss.metrics.repl.apply.ops, opCount + offset, "wrong number of applied ops"); diff --git a/src/mongo/db/repl/sync_tail.cpp b/src/mongo/db/repl/sync_tail.cpp index a5dfbf3ebb4..306e0b4185c 100644 --- a/src/mongo/db/repl/sync_tail.cpp +++ b/src/mongo/db/repl/sync_tail.cpp @@ -147,6 +147,11 @@ public: Counter64 opsAppliedStats; ServerStatusMetricField<Counter64> displayOpsApplied("repl.apply.ops", &opsAppliedStats); +// Tracks the oplog application batch size. +Counter64 oplogApplicationBatchSize; +ServerStatusMetricField<Counter64> displayOplogApplicationBatchSize("repl.apply.batchSize", + &oplogApplicationBatchSize); + // Number of times we tried to go live as a secondary. Counter64 attemptsToBecomeSecondary; ServerStatusMetricField<Counter64> displayAttemptsToBecomeSecondary( @@ -350,8 +355,10 @@ Status SyncTail::syncApply(OperationContext* opCtx, bool isNoOp = opType[0] == 'n'; if (isNoOp || (opType[0] == 'i' && nss.isSystemDotIndexes())) { - if (isNoOp && nss.db() == "") + if (isNoOp && nss.db() == "") { + incrementOpsAppliedStats(); return Status::OK(); + } Lock::DBLock dbLock(opCtx, nss.db(), MODE_X); OldClientContext ctx(opCtx, nss.ns()); return applyOp(ctx.db()); @@ -1495,6 +1502,9 @@ StatusWith<OpTime> multiApply(OperationContext* opCtx, "attempting to replicate ops while primary"}; } + // Increment the batch size stat. + oplogApplicationBatchSize.increment(ops.size()); + std::vector<Status> statusVector(workerPool->getNumThreads(), Status::OK()); { // Each node records cumulative batch application stats for itself using this timer. |