Redis: don't record values that are written to the DB #50
Conversation
Redis instrumentation now skips recording actual values written to the database for known write commands.
| argsLength := len(parts) - 1 | ||
| numArgs, ok := cmdCaptureArguments[cmd] | ||
| if !ok { | ||
| return cmdStr, cmd, argsLength |
There was a problem hiding this comment.
If I understand this logic correctly, any unrecognized/not-explicitly-encoded command will just have part[0] (the command name and only the command name, no args) as the span name?
There was a problem hiding this comment.
It's the reverse. Any unrecognized command will use the full command with args as the tag value (preserve current behavior).
There was a problem hiding this comment.
Perhaps for safety it should be the other way? The Jedis (Java) instrumentation appears to just use the command name (first part) and nothing else (though I like the idea of porting the "this number of params is safe" there).
There was a problem hiding this comment.
The list of known commands is quite exhaustive. I actually went over all commands and listed down every command I found that wrote data to the DB or queried data using values that might be in the DB. All other commands either use keys/IDs and never take any "data" as arguments so it's safe to keep them at least from data volume perspective. The downside is that a new "write" command in future would escape this filter but it's probably not that big of a deal given than this PR mainly tries to reduce trace volume.
Redis instrumentation now skips recording actual values written to the
database for known write commands.