Running joinerCheck for all databases

Overview

The joinerCheck program is run bi-monthly on all active databases on the hgwbeta machine. Errors that are found may be newly introduced or pre-existing. Corrections to data and/or all.joiner rules that impact browser users are given priority; corrections to known errors are made as time permits.

• When/Who: Run bi-monthly by a member of QA
• Why: Catch global rule problems with all.joiner or new data problems within assemblies
• Where: hgwbeta:/hive/groups/qa/joinerCheck
• Remember: Before starting, be sure to update your source tree so that you have the latest version of all.joiner

Run joinerCheck -keys

• Create new directory at: /hive/groups/qa/joinerCheck/keys/YYYYMM
• In the past it has been useful to run against all DBs in three batches: (1) human, (2) mouse and rat, and (3) the rest. One batch per day, for three days in a row
• The goal is to have each DB run without conflicting with genbank updates. Never run overnight on a Sunday (EST updates). In the past, running Mon-Tue-Wed has been successful, starting in the morning after the previous night's genbank update has finished (~8am) and ending the batch before the next night's genbank updates start (~12pm)
• Execute Keys program hgwbeta:/hive/groups/qa/joinerCheck/keys/YYYYMM> nice nohup joinerBatchKeys.csh [hg | mm_rn | other | extra] &
• When the script finished, it sends an email to whoever ran it.
• When complete, execute Error program> getJoinerKeyErrors.pl /hive/groups/qa/joinerCheck/keys/YYYYMM

Evaluate -keys result

• Create working directory to investigate problems at: /hive/groups/qa/joinerCheck/rules/YYYYMM
• Review results of getJoinerKeyErrors.pl summary report
• Compare to the last results to find out what is known, what is new, and what has been fixed since the last run
• Analysis hints:
  • Review makedocs and all.joiner for tracks/assemblies with issues
  • Attempt to isolate exact problem. Is it data or a rule problem?
  • Original developer can help confirm track problems
  • If it is a genbank table, confirm that genbank updates are on then contact Mark Diekhans
  • Data fixes can often be made by QA, rule fixes are usually made by a developer

Run joinerCheck -times

• Create new directory at: /hive/groups/qa/joinerCheck/times/YYYYMM
• Execute Times program hgwbeta:/hive/groups/qa/joinerCheck/keys/YYYYMM> nice nohup joinerBatchTimes.csh [hg | mm_rn | other | extra] &

Evaluate -times result

• Use -keys working directory to investigate problems at: /hive/groups/qa/joinerCheck/rules/YYYYMM
• Review results
• Compare to the last results to find out what is known, what is new, and what has been fixed since the last run
• Analysis hints:
  • Any and all output is considered an error
  • Priority of found errors can be high if they involve proteome/known genes data
  • Makedoc notes are the primary source for tracking changes
  • Analysis hints from -keys analysis can be helpful for isolating root cause of -times errors, too

• If you need to touch a table's timestamp, do this:

> sudo mytouch database table-pattern [touchtime]

NOTE: use single quotes around table-pattern if it contains shell special chars such as * or ?

NOTE: use the syntax of man touch for touchtime

Example: [[CC]YY]MMDDhhmm[.ss]

When we use this is QA, we always set the time like so:

all.joiner rule: table1 must be created before table2

reality: table2 was created before table1

use mytouch: touch table2 so that its time and date are the same as table1 plus the next full hour with 00 minutes and 00 seconds.

EXAMPLE: table2 updateTime = 2006-10-23 08:30:16 table1 updateTime = 2006-10-23 13:04:22

> sudo mytouch database table2 200610231400.00

Also, be sure to edit the make doc of the assembly for the database table you are changing. Make an edit in the section in which the table was originally documented like so:

QA NOTE: [your name, today's date] mytouch to table2 200610231400.00 (because it was created after table1).