Details
-
Bug
-
Resolution: Fixed
-
Major - P3
-
None
-
None
-
None
-
8
-
Storage - Ra 2020-06-15, Storage - Ra 2020-06-29
Description
We have recently (within the last week) seen two different symptoms indicating that the content of documents in MongoDB are getting corrupted.
We should understand what could be causing this - most likely a change to how WiredTiger handles modify records, and fix the problem. It's possible that this failure is outside of WiredTiger but unlikely.
The symptoms are:
When running:
jstests/hooks/run_validate_collections_background.js:182:19
|
The following error is reported:
[ReplicaSetFixture:job0:secondary1] | 2020-06-09T18:35:20.607+00:00 I STORAGE 4835002 [conn7424] "Document corruption details - Document validation failure; size mismatch","attr":{"recordId":{"RecordId":1},"validatedBytes":18,"recordBytes":37}
|
And also:
After running:
jstests/concurrency/fsm_workloads/agg_out.js
|
As part of a concurrency test suite (i.e: interleaved with other test cases), a validation fails with:
[ReplicaSetFixture:job0:secondary0] | 2020-06-06T01:15:46.171-07:00 I STORAGE 4835000 [conn1848] "Document corruption details - Multiple causes for document validation failure; error status and size mismatch","attr":{"recordId":{"RecordId":1},"validatedBytes":0,"recordBytes":147,"error":{"code":22,"codeName":"InvalidBSON","errmsg":"bson length doesn't match what we found in object with _id: 2.0"}}
|
Attachments
Issue Links
- is related to
-
WT-6418 Account for aborted updates when doing the first scan of updates before inserting them to the history store
-
- Closed
-