Terminating process due to non-recoverable failure

[RanaHossainUK]

Hi - we are using Esent with heavy load, and experiencing following issues; first one is a crash that takes down the whole process:

• XXX.Service (4624,G,0) Terminating process due to non-recoverable failure: PV: 6.2.9200.0 SV: 6.2.9200.0 GLE: 0 ERR: 0(:0): bf.cxx(25721) (ESENT[6.2.9200.0] RETAIL RTM MBCS). Tag: xechkCheckPre == ppghdrPost->checksum.

• XXX.Service (25496,D,0) cc499a7cc368408aab616e93c683a844: A request to write to the file "C:\Users\Administrator\Desktop\loader2\db\db.edb" at offset 243990528 (0x000000000e8b0000) for 98304 (0x00018000) bytes has not completed for 36 second(s). This problem is likely due to faulty hardware. Please contact your hardware vendor for further assistance diagnosing the problem.

To try to narrow down where these errors may be coming from: we created a LoadSimulator with only one entity (DataEntry) which contains 4 guids (one being primary), 3 additional indexes (one being a multiple property index), and 26 other properties (mostyle double, int, bool and one being a byte-array).

Finally, we created some dummy data and running some tasks that would keep running (repeat once completed: each insert will have a new guid for PK): we also run multiple instance of task based on processor count.

1. Read list of items using seek and start key values with a delay of 500ms between read
2. Single insert a list of 1M items with a delay of 10ms between items
3. Single LAZY insert a list of 1M items with a delay of 10ms between items
4. Batch insert 1K with Lazy flush from a list of 1M items with a delay of 10_000ms between each batch
5. Batch insert 1K with CommitType.None from a list of 1M items with a delay of 1000ms between each batch

If we run task 3 (SingleInsertLazy) and task 5 (BatchInsertLazy), whole process crashes (usually within couple of minutes) (on a pc with 8 cores, each task is duplicated so 4 instances for each task)

If we run task 2 (SingleInsert NonLazy) and task 4 (BatchInsert NonLazy), we get the error of "abnormally long time" (usually within 10 to 15 minutes) (on a pc with 8 cores, each task is duplicated so 4 instances for each task)

Any ideas or suggestions?