INTPYTHON-309 & INTPYTHON-417 Use new cluster and schedule on interval #50
Conversation
… into INTPYTHON-309
|
Before merging, we need to update the Evergreen project config to only run the local variants on pull requests. |
.evergreen/config.yml
Outdated
| tasks: | ||
| - name: test-semantic-kernel-python | ||
| - name: test-semantic-kernel-python-local | ||
| commands: | ||
| - func: "fetch repo" | ||
| - func: "setup local atlas" | ||
| - func: "execute tests" | ||
|
|
||
| - name: test-semantic-kernel-python-remote | ||
| commands: | ||
| - func: "fetch repo" | ||
| - func: "setup remote atlas" | ||
| - func: "execute tests" | ||
|
|
||
| - name: test-semantic-kernel-csharp-local | ||
| commands: | ||
| - func: "fetch repo" | ||
| - func: "setup local atlas" | ||
| - func: "execute tests" | ||
|
|
||
| - name: test-semantic-kernel-csharp-remote | ||
| commands: | ||
| - func: "fetch repo" | ||
| - func: "setup remote atlas" | ||
| - func: "execute tests" | ||
|
|
||
| - name: test-langchain-python-local |
There was a problem hiding this comment.
Out of scope for this ticket.
We should make a python generator for this as well. That or change this to just two tasks.
local run and remote run
There's no difference in the commands otherwise.
There was a problem hiding this comment.
Lol, replied to the wrong comment: https://jira.mongodb.org/browse/INTPYTHON-441
| - func: "execute tests" | ||
|
|
||
| - name: test-semantic-kernel-csharp | ||
| - name: test-langchain-python-remote |
There was a problem hiding this comment.
Do we need to add tags or can we do regex matching in evergreen to make sure these aren't run on pull requests
.evergreen/setup-remote.sh
Outdated
| # Ensure the remote database is populated. | ||
| . .evergreen/utils.sh | ||
|
|
||
| PYTHON_BINARY=$(find_python3) | ||
|
|
||
| # Should be called from src | ||
| EVERGREEN_PATH=$(pwd)/.evergreen | ||
| TARGET_DIR=$(pwd)/$DIR | ||
| SCAFFOLD_SCRIPT=$EVERGREEN_PATH/scaffold_atlas.py | ||
|
|
||
| mkdir atlas | ||
|
|
||
| pushd atlas | ||
|
|
||
| $PYTHON_BINARY -m venv . | ||
| source ./bin/activate | ||
| popd | ||
|
|
||
| # Test server is up | ||
| $PYTHON_BINARY -m pip install pymongo | ||
| CONN_STRING=$MONGODB_URI \ | ||
| $PYTHON_BINARY -c "from pymongo import MongoClient; import os; MongoClient(os.environ['MONGODB_URI']).db.command('ping')" | ||
|
|
||
| # Add database and index configurations | ||
| DATABASE=$DATABASE \ | ||
| CONN_STRING=$MONGODB_URI \ | ||
| REPO_NAME=$REPO_NAME \ | ||
| DIR=$DIR \ | ||
| TARGET_DIR=$TARGET_DIR \ | ||
| $PYTHON_BINARY $SCAFFOLD_SCRIPT |
There was a problem hiding this comment.
This could call provision-atlas as a next step and remove duplication.
| "numDimensions": 1536, | ||
| "path": "embedding", | ||
| "similarity": "cosine", | ||
| "type": "vector" |
There was a problem hiding this comment.
This matches the value in the previous Atlas cluster:
| if collection_name in collections: | ||
| logger.debug("Clearing existing collection", collection_name) | ||
| db[collection_name].delete_many({}) |
There was a problem hiding this comment.
NIT: We could also drop the collection entirely. This would remove all existing index definitions on it.
There was a problem hiding this comment.
I tried that but it seemed to cause race conditions.
| logger.debug( | ||
| "creating search index: %s on %s.%s...", | ||
| index_name, | ||
| database_name, | ||
| collection_name, | ||
| ) |
| semantic-kernel-python) | ||
| MONGODB_URI=$SEMANTIC_KERNEL_MONGODB_URI | ||
| ;; | ||
| semantic-kernel-csharp) | ||
| MONGODB_URI=$SEMANTIC_KERNEL_MONGODB_URI |
There was a problem hiding this comment.
Out of scope for this PR:
Could it be that sharing the URI could be leading to some failures in index definitions?
There was a problem hiding this comment.
I don't think so, given they're using different DBs.
| scaffold_atlas() { | ||
| PYTHON_BINARY=$(find_python3) | ||
|
|
||
| # Should be called from src | ||
| EVERGREEN_PATH=$(pwd)/.evergreen | ||
| TARGET_DIR=$(pwd)/$DIR | ||
| SCAFFOLD_SCRIPT=$EVERGREEN_PATH/scaffold_atlas.py | ||
|
|
||
| mkdir -p atlas | ||
| pushd atlas | ||
|
|
||
| $PYTHON_BINARY -m venv . | ||
| source ./bin/activate | ||
| popd | ||
|
|
||
| # Test server is up | ||
| $PYTHON_BINARY -m pip install pymongo | ||
| CONN_STRING=$CONN_STRING \ | ||
| $PYTHON_BINARY -c "from pymongo import MongoClient; import os; MongoClient(os.environ['CONN_STRING']).db.command('ping')" | ||
|
|
||
| # Add database and index configurations | ||
| DATABASE=$DATABASE \ | ||
| CONN_STRING=$CONN_STRING \ | ||
| REPO_NAME=$REPO_NAME \ | ||
| DIR=$DIR \ | ||
| DEBUG="${DEBUG:-1}" \ | ||
| TARGET_DIR=$TARGET_DIR \ | ||
| $PYTHON_BINARY $SCAFFOLD_SCRIPT | ||
| } |
Also resolves INTPYTHON-313 by adding support for docarray on local Atlas.