[OOZIE-1980] Sql error should not fail coord job

• Type: Bug
• Status: Patch Available
• Priority: Major
• Resolution: Unresolved
• Assignee: Purshotam Shah
• Reporter: Purshotam Shah
• Created: 26/Aug/14 06:55
• Updated: 06/Feb/18 17:26

Description

We know that sql error is transient, it should not fail the job.
We should ignore the SQL error, so that next materialization cycle can pick the job.

protected void materialize() throws CommandException {
        Instrumentation.Cron cron = new Instrumentation.Cron();
        cron.start();
        try {
            materializeActions(false);
            updateJobMaterializeInfo(coordJob);
        }
        catch (CommandException ex) {
            LOG.warn("Exception occurred:" + ex.getMessage() + " Making the job failed ", ex);
            coordJob.setStatus(Job.Status.FAILED);
            coordJob.resetPending();
            // remove any materialized actions and slaEvents
            insertList.clear();
        }
        catch (Exception e) {
            LOG.error("Exception occurred:" + e.getMessage() + " Making the job failed ", e);
            coordJob.setStatus(Job.Status.FAILED);
            try {
                CoordJobQueryExecutor.getInstance().executeUpdate(CoordJobQuery.UPDATE_COORD_JOB_MATERIALIZE, coordJob);
            }
            catch (JPAExecutorException jex) {
                throw new CommandException(ErrorCode.E1011, jex);
            }
            throw new CommandException(ErrorCode.E1012, e.getMessage(), e);
        } finally {
            cron.stop();
            instrumentation.addCron(INSTRUMENTATION_GROUP, getName() + ".materialize", cron);
        }
    }

Issue Links

• relates to: OOZIE-3134 Potential inconsistency between the in-memory SLA map and the Oozie database