ETL troubleshooting recipes (To make changes permanent in DB don't forget to issue COMMIT at the end.)

Issue: When upstream data change, some BSP plating info in BSP_PLATED_SAMPLES might need to be refreshed.

Remedy:  Request a refresh for BspPlatingRequestAgent by issuing SQL like this in GAP_PROD.ANALYTICS:

INSERT INTObsp_plating_special_refreshesVALUES('6E3K2', SYSDATE);

COMMIT;

Note: There is no "SM-" prefix,

Issue: When upstream data change, some samples in PDO_STAR5 might need to be refreshed.

Remedy: Follow the instructions here for Manual refresh of PDO Star5 using a query. The query should have the format below:

SELECT DISTINCT a.pdo_name||','||a.pdo_sample_id FROM ... a WHERE ...

Issue/Error: Once a week (1am on Fri) PdoStar5 goes through 1y-refresh. Sometime you might want to trigger it earlier.

Diagnosis: 

Remedy:  Connect to "analytics" server using your favorite SSH tool and run the following line:

/home/unix/analytics/TigerETL3/runEtlAgent.sh Task task=etlObject agentName=analytics.tiger.agents.PdoStar db=analyticsetl 'delta=MillisDelta.loadFromDb(365d,0d)'

Issue: How to make PDO_STAR5 ETL pick a new product ?

Remedy: Add a new "interesting" product by issuing SQL like this in SEQPROD.COGNOS. Provide part number.

INSERT INTO pdo_star_interesting_products values('P-123456');

Note: Also create new JIRA RPT ticket and assign to raymond to track the following:  Assign PdM-approved minimum requirement(s) for coverage met.  pdo_star5_aux will need product_goal assigned as well as sample_coverage_normalized and sample_coverage_unformatted set.  DCFM_RULES must be updated.  DCFM_Agent may need to be updated.  PDO_SEQ_STAR needs "Sample Coverage" calculated field updated.

Issue:Missing BAM file path in PDO Tracker or WGS/WES Project Tracker report (PDO_STAR5_Aux DM) (for on-prem only)

Cause: 

1. Seq Only PDOs do not get BAM files in PDO_STAR5_Aux DM
2. Plated sample in BSP_Plated_Samples is different than slxre_readgroup_metadata.sample_id
3. Read group of the PDO Sample is missing in SLXRE2_RGHQS, most likely because there isn't a record in metrics.basecall_analysis for metrics_type Indexed or Unindexed.
4. PM billed sample before PDOSTAR5 picked the most recent aggregation or before sample met full coverage

Remedy: 

1. None
2. Check if Mercury DWH and Mercury Run API still do not agree about the plated sample. If they don't - seek help from Mercury team. If they match, a refresh of the flowcell should correct it. BSP exported samples might be associated with the wrong PDO in BSP_Plated_samples because the mapping in BSP is not accurate. If it is important to fix the data , BSP team could help and then request a refresh for the affected plated sample as instructed in the recipe on this page to update BSP Plating DM. Finally, make sure to refresh in PDO Star5 if it won't happen automatically.
3. Make ticket for the PIpeline and refer to PO-13231 as being the same issue.
4. PDOs created within past year will get refreshed on Friday 2:55 am or Analytics manually refresh for faster data delivery. PDOs older than 1 year MUST be manually refreshed.

Issue: When upstream data change, some samples in ReadgroupMetadata might need to be refreshed.

Remedy: Request a refresh for RunMercuryAgent by issuing SQL like this in SEQPROD.COGNOS:

BEGIN

 cognos.data_warehouse_api.request_refresh(P_FLOWCELL_BARCODE=>'XYZ', P_REQUESTED_BY=>'JohnDoe', P_COMMENTS=>'Test', P_DEFERRAL_ANALYSIS_DATE=>null);

END;

Note: No COMMIT needed in this particular call - it's implicitly done for you. Just replace the XYZ with a flowcell barcode. Leave all other parameters there and unchanged.

Issue: How to prevent an entire FC from getting into ReadgroupMetadata (Analytics Blacklisting).

Remedy: Follow the instructions here for blacklist runfolder from RunMercury/RghqsLoader agents

Note: Blacklist_timestamp is not null for blacklisted flowcells.

Issue: How to whitelist a FC for processing by Run Mercury and RGHQS agents.

Remedy: Follow the instructions  here for whitelist runfolder from RunMercury/RghqsLoader agents

Issue: How to refresh Picard Aggregation DMs for a specific set of samples. Note, this will not refresh any metrics unless they are part of a new aggregation (new version). Only basecall metrics rolled up at library and sample level get updated

Remedy: Follow the instructions  here for Manual Refresh of Picard Aggregations for a list of samples.

Issue: PicardAggregator ETL fails with (COGNOS.SLXRE2_PAGG_SAMPLE_PK) violated error.

Cause: Most likely Is_latest flag is set for more than 1 aggregations.

Remedy: Find the sample that cause it (run the query below) and make a blocker PO ticket + notify somebody from Pipeline team to address it ASAP because the ETL won't recover until this is fixed.

...

Issue: Designation ETL fails with "ORA-30926: unable to get a stable set of rows in the source tables" error

Diagnosis: This is usually caused by a flowcell with 2 runs valid run (uncanceled), one of which should have been cancelled. Run the query below to determine the trouble flowcell:

...

Remedy: Contact Tammy, Andrew B. , Mike DaSilva or John W. to clear the bad run from their dashboard. Designation ETL will recover as soon as the run is cleared. 

Issue: Scala ETL ERROR: analytics.etl.DeckEventAgent$ (Most likely 'lcSet' tag is missing)

Diagnosis: Sometimes this Squid web-service returns no LcSet. Comment about this incident is automatically posted inIPI-61224and ticket reopened.

Remedy: Just follow up with LIMS team. OurDeckEventAgent will catch up as soon as LIMS service is fixed. No other intervention is needed.

Issue: analytics.etl.PkViolationWarning. unique constraint (COGNOS.RGHQS_METADATA_PK) violated

Diagnosis: This happens when 2+ readgroups having same (fc, lane, mol_index) try to get into datamart. Picard calls it "CLASH". RunMercuryAgent would ignore abusing readgroups and move on with the rest of 'good' ones. Email would give you detailed troubleshooting info.

Remedy: There isn't much for us to do. If upstream issues are resolved you can request a ReadgroupMetadata refresh for the whole run/fc (see above). You can search for the RG in GPINFOJIRA by searching the FC ID to see if it has been blacklisted and already handled upstream.

Issue/Error: ETL: DB error, server: (thorium or vorarestore), Error: ORA-12008: error in materialized view refresh path, ORA-25153: Temporary Tablespace is Empty

Diagnosis: This happens when the DB team is doing some kind of backup or transfer.

Remedy: These errors can be ignored. Everything is actually fine.

...