throughput_report fails for large number of events

[cms21]

throughput_report fails for a large number (512 or greater) number of events. It fails in its request to the server because of a too long URL. Possibly need a different type of HTTP call. Sample error below:

Traceback (most recent call last): File "analytics.py", line 14, in <module> times, done_counts = throughput_report(events, to_state="JOB_FINISHED") File "/home/csimpson/.local/miniconda-3/latest/lib/python3.7/site-packages/balsam/analytics.py", line 14, in throughput_report times = list(log.timestamp for log in finished_logs) File "/home/csimpson/.local/miniconda-3/latest/lib/python3.7/site-packages/balsam/_api/query.py", line 114, in __iter__ self._fetch_cache() File "/home/csimpson/.local/miniconda-3/latest/lib/python3.7/site-packages/balsam/_api/query.py", line 128, in _fetch_cache offset=self._offset, File "/home/csimpson/.local/miniconda-3/latest/lib/python3.7/site-packages/balsam/_api/manager.py", line 204, in _get_list count, results = self._fetch_pages(filter_chunk, ordering, limit, offset) File "/home/csimpson/.local/miniconda-3/latest/lib/python3.7/site-packages/balsam/_api/manager.py", line 170, in _fetch_pages response_data = self._client.get(self._api_path, **query_params) File "/home/csimpson/.local/miniconda-3/latest/lib/python3.7/site-packages/balsam/client/rest_base_client.py", line 65, in get return self.request(url, "GET", params=kwargs) File "/home/csimpson/.local/miniconda-3/latest/lib/python3.7/site-packages/balsam/client/requests_client.py", line 98, in request response = self._do_request(absolute_url, http_method, params, json, data) File "/home/csimpson/.local/miniconda-3/latest/lib/python3.7/site-packages/balsam/client/requests_client.py", line 130, in _do_request self._raise_with_explanation(response) File "/home/csimpson/.local/miniconda-3/latest/lib/python3.7/site-packages/balsam/client/requests_client.py", line 150, in _raise_with_explanation response.raise_for_status() File "/home/csimpson/.local/miniconda-3/latest/lib/python3.7/site-packages/requests/models.py", line 1021, in raise_for_status raise HTTPError(http_error_msg, response=self) requests.exceptions.HTTPError: 414 Client Error: URI Too Long for url: https://balsam-dev.alcf.anl.gov/events/?job_id=29682869&job_id=29682870&job_id=29682871&job_id=29

[masalim2]

This should be something we can fix by decreasing the FILTER_CHUNK_SIZE parameter slightly in the client-side manager.py module:

balsam/balsam/_api/manager.py

Line 13 in da4e521

FILTER_CHUNK_SIZE = 512

The clients are already using this parameter to split up a large request into chunks, make the requests separately, and then stitch the responses back together. We might be running up against a new limit here with larger job_id keys 😄

[masalim2]

A more flexible solution might be to check for the 414 URI Too Long error specifically, and if that is thrown, retry the request with smaller filter chunk size. Otherwise, I agree that a different HTTP call that keeps the long list of job_ids out of the URL (using the request body instead) would be reasonable