This summary was written by AI and I do not endorse it:

Summary

Wire retry-delay recording at the timeseries retry-detection call sites for SERVER-107365 (retried writes latency histogram).

Sites

src/mongo/db/timeseries/write_ops/timeseries_write_ops.cpp (~L102) command-level

src/mongo/db/timeseries/write_ops/internal/timeseries_write_ops_internal.cpp (~L180) statement-level

On each retry branch, compute

millis = Date_t::now() - originalWriteWallClock

and call

RetryableWritesStats::get(opCtx)->recordRetriedWriteDelay(millis)

alongside incrementRetriedStatementsCount().

Note

Confirm an original-write timestamp is available at each of these sites (opTime or table OplogEntry); if not available at a given site, document and defer that one.

Test

Add one test asserting a retried timeseries write lands one sample in the expected bucket (SERVER-107365 Task 3: one test per call site).

References

• SERVER-107365 (Add Metrics for Retried Writes)
• Prereqs: SERVER-134421 (metric definition), SERVER-134422 (record helper)