Fix extraction failure - Githubissues

fkengun commented 2 weeks ago

Changes in this PR (address issue #166 as well):

Move serialization buffer from member variable to thread-local in StoryChunkExtractorRDMA
Move extraction RPC call into try body to make sure Cereal and stream object are still in scope
Catch more exceptions in deserializing story chunk
Remove dependency on Cereal in external_libs
Increase MAX_BULK_MEM_SIZE to 4MB

fkengun commented 2 weeks ago

Without any errors/exceptions during extraction, the previously missing lines are also fixed now. All the events from client_lib_multi_storytellers are saved to csv files.

EnekoGonzalez3 commented 1 week ago

I have run the local script with the updates on this PR. With this changes: 3/4 files are generated, 1/3 generated files are complete.

Previous to this changes: 3/4 files are generated, 2/3 generated files are complete.

EnekoGonzalez3 commented 4 days ago

I am still experimenting lot of inconsistency. I have been able to have some successful tests, but most of them fail by different reasons. In any case, this is one of the errors I have found on the grapher log: [2024-06-26 17:40:50.506] [ChronoGrapher] [error] [GrapherRecordingService] Failed to deserialize a story chunk, ThreadID=6. Cereal exception encountered. [2024-06-26 17:40:50.506] [ChronoGrapher] [error] [GrapherRecordingService] Exception: Failed to read 7 bytes from input stream! Read 5

Also, after this changes I have some executions that never end. The stop process is not able to finish the execution. It happens both on keeper level and grapher level.

grc-iit / ChronoLog

Fix extraction failure #175