[cassandra] Fix batched query leak #714
Conversation
brettlangdon
left a comment
brettlangdon
left a comment
There was a problem hiding this comment.
I assume you tested this with agent obfuscation as well?
|
Yes this was tested with the agent obfuscation. However your comment made me check again and I caught an issue where including the query as a tag for a cassandra span will not result in the query being overwritten with the obfuscated query. I've updated the PR description with the before and after of a trace. One more thing to consider is if there are issues with using multiple combined queries as a resource. |
|
Also, have you tested with running multiple queries at once in batch? Either 2-3 of the same query or 2-3 of different queries. do we handle obfuscation properly there? Other than that and the one question I had re: assertion, this appears to be mostly good to go. |
|
Obfuscation seems to work with multiple queries, including different types of queries. |
|
One last small question, but otherwise this is good to go. |
|
There is a merge conflict here now as well. |
| # reset query if a string is available | ||
| resource = getattr(query, "query_string", query) | ||
| elif t == 'BatchStatement': | ||
| resource = 'BatchStatement' |
This reverts commit cb245cd.
The cassandra integration does not adhere spans created for batch statements to the specification required by the agent to obfuscate the query.
This PR changes the logic so that query is set as the resource of the span. Since the span type is already set to
'cassandra'the query can now be properly obfuscated by the agent.Before:
After:
