DRIVERS-1677, DRIVERS-1673: Add general logging specification and command logging

[kmahar]

This is the first in a series of PRs for the logging spec.

This PR adds:

• A new general logging specification describing requirements for how drivers implement logging and defining shared concepts other specs can use, e.g. common log levels and component names
• Additions to the unified test format to support logging assertions (I know there is an existing PR that also adds a schema version 1.11; if that goes in first I will update to use 1.12 instead).
• Rebrands the command monitoring spec as the command logging and monitoring spec (I considered CMAL but it felt like the acronym CLAM was too good to pass up...), and adds information about command logging, along with both unified format and prose tests.

Future PRs will add logging and tests for CMAP, SDAM, and server selection.

[jyemin]

Lots of comments, but this looks fantastic. Nice work.

[jyemin]

Then -> When

[kmahar]

thanks; ended up deleting this per above thread about using a single boolean

[kmahar]

I'm confused about the "Check if all JSON test files are up-to-date" failure. it seems to be complaining about a CMAP test I haven't touched here, and the "diff" is a line that looks identical before and after. it doesn't seem relevant

diff --git a/source/connection-monitoring-and-pooling/tests/pool-clear-interrupt-immediately.json b/source/connection-monitoring-and-pooling/tests/pool-clear-interrupt-immediately.json
index e61ca51..54e2566 100644
--- a/source/connection-monitoring-and-pooling/tests/pool-clear-interrupt-immediately.json
+++ b/source/connection-monitoring-and-pooling/tests/pool-clear-interrupt-immediately.json
@@ -24,7 +24,7 @@
       "name": "waitForEvent",
       "event": "ConnectionPoolCleared",
       "count": 1,
-      "timeout": 1000      
+      "timeout": 1000
     },
     {
       "name": "waitForEvent",

edit: ah it's a whitespace change. I'll make a small separate PR to correct that so we can get a complete CI run on the new tests. I wonder why this only showed up now... the js-yaml version hasn't changed

[jmikola]

I didn't expect this would require a subsequent assertion on the contents of data field in the test file. Rather, the mere presence of failureIsRedacted would require a driver to assert that there is some failure to inspect in the first place. In other words, a test could use failureIsRedacted on its own with no meaningful data assertion and trust the assertion would fail outright if the log data omitted a failure altogether.

I imagine there might also be cases where data.failure could be something other than { $$exists: true } (perhaps a $$type assertion or something more specific.

[kmahar]

ah, I misunderstood your comment above. and good point re: other assertions on failure being possible.

I updated the language here to just say

This field SHOULD only be specified when the log message data is expected to contain a failure value.

[kmahar]

#1316 bumped the unified format schema version; just pushed another few commits to reflect this so that the logging changes are added via schema version 1.13.

[jyemin]

LGTM!!!!

[BorisDog]

LGTM (2 minor comments)

[BorisDog]

nit: redundant space in reply .?

[kmahar]

removed. thanks!

[BorisDog]

For your consideration, you could add a link to the truncation section #truncation-of-large-documents.

[kmahar]

good thought to add a link. that section covers truncation rationale, but I am going to add one to the section on truncation rules: Configurable Max Document Length.