pika-server 3.0
UTC timezone
Server only works with unix timestamps in seconds. Server UTC timezone does not matter!
-
Server does not provide the client with an utc_timezone -
Client can send:
a) "format_result": {"utc_timezone": "+01:00"} --> Server returns formatted time in client utc_timezone
b) "format_result": true --> Server returns formatted time in utc_timezone "00:00"
c) "format_result": false
-
If server needs to determine current time, use: import time; int(time.time()) -
Remove translate_timezone (not needed anymore), get_utc_timezone and format_time have to use client utc_timezone -
Test client in different timezones and run pika-server in docker container, see pika-packages/pika-server
Further changes
-
refactor metric.py (use class, it is called twice at the moment) -
start_time and end_time should return formatted time if requested (all other metrics from job_data should also return 'n/a' if None etc.) --> start_unixtime needed for GUI -
remove leading units (also zeros) from duration metrics if only one dataset is provided (/job/jon_id...) -
check the duration between data blocks in data collection (if no data is received within the last hour, raise CRITICAL) -
Unit tests need to be adjusted (currently commented out) -
Write unit tests for new endpoints /node, /reservation, and /job_pend -
use_date_interval in query_filter -
/footprint should also use query_filter -
/footprint is using a different orm model? (hard coded column names) -
Replace live and not_live with live: boolean (live=true --> running, live=false --> not running) -
Endpoint for tags /tag (returns job_tag table, similar to /partition) -
query_settings instead of query_filter in src/rest_api, see /job or /timeline -
server send current version to client that client can make compatibility check -
rename database tables (Frank)
Job_Data --> job_data
Job_Tag --> job_tag
new column names according to job_queue
-
server adjustments according to new database tables and columns -
client adjustments according to new database tables and columns (Frank) -
global get_formatted_results() function that can be used by all views (see: src/sql/view/node/nody.py) -
metric.py simpler format for io (maybe generate metric.py at start time, adding a new io system is horrible at the moment) -
pending_time not working for pending jobs -
remove sql table partition_table from database and use partition.json with the same information