GitLab

schema_change: detect and register RELOCATED events that used to mess up things on LSF 9.0 ("job rescheduled" in LSF parlance)

ResourceDescription expanded to include both submission_cmd_args and worker_cmd_args. Both args can be specified in a PipeConfig file.

schema change: turned all VARCHAR(<255) into VARCHAR(255) -- should improve experience with long host namest (thanks, MichaelP!)

Updated schema diagram and description file that include param_id_stack and accu_id_stack in job table.

"parameter stack" implementation using two extra fields in job table. Accu content intended for any job_id has preference over Input_id content for the same job.

Schema change: changed the data type of monitor.analysis to TEXT as per Michael Paulini's suggestion, to fit more and longer analysis names.

use SqlSchemaAdaptor to detect the current code's sql version and record it in 'hive_meta' (leave it out of tables.*sql* files)

move Core 'schema_version' out of tables.* files into HiveGeneric_conf (via ApiVersion), expose it for manipulation and make it available to PipeConfigs

new 'hive_meta' table to keep hive_sql_schema_version (=number of patches), hive_pipeline_name and hive_use_triggers

Has one known limitation (always INSERTs instead of INSERT_IGNORE and REPLACE) and may contain other bugs.

Copied the @-tag annotation from tables.sql to tables.sqlite. Unlike the original mysql version, the SQLite version gives no warnings when processed by sql2html.pl

change of default behaviour: hive_capacity is now off by default (=NULL); setting hive_capacity=0 or analysis_capacity=0 stops scheduling AND specialization to a particular analysis

added a new per-analysis "analysis_capacity" limiter for cases where users want to limit analyses independently

Introduced a new 'NO_ROLE' cause_of_death for failures during specialization (not so much of an error, really!)

capture Worker's death message during the new 'SPECIALIZATION' status in job_message/msg (thanks, Thomas!)

feature: jobless workers will now leave module compilation errors in the job_message table (thanks, Kathryn!)

cause_of_death="" no longer used for decision making, cause_of_death IS NULL by default and FATALITY renamed UNKNOWN for clarity

introduced semaphored_job_count, renamed unclaimed_job_count-->ready_job_count, changed reporting, fixed hive_capacity=0

moved failed_job_tolerance, max_retry_count, can_be_empty and priority columns from analysis_stats to analysis_base

added specific defaults into analysis_stats and analysis_stats_monitor; re-ordered the fields for easier navigation

Dropped 'BLOCKED' job status and introduced 'SEMAPHORED' status that is maintained in sync with semaphore_counts; less confusing and more efficient (with new 3-part index)

record each Workers log_dir in the database; simplified the log_dir code and renamed cmdline options accordingly

At last rename GET_INPUT into FETCH_INPUT for consistency between the schema and the code (it seems to be harder to patch all the accumulated code)

replaced rc_id by resource_class_id throughout the schema and added the foreign keys on resource_class_id