Closed Bug 574753 Opened 16 years ago Closed 12 years ago

processors need to be able to work around isolated errors interacting with hbase

Categories

(Socorro :: General, task)

task
Not set
normal

Tracking

(Not tracked)

RESOLVED FIXED

People

(Reporter: dre, Assigned: lars)

Details

If a region is down and that causes one or two items to be inaccessible, the processor should set them aside in an exceptions table and continue working.
More info: The current behavior for a processor is to retry, then fail, then die. Then the next processor tries to pick up the next job from the monitor but it's the same job, same problem, next processor dies. Repeat until we don't have any processors.
Assignee: nobody → lars
Severity: normal → critical
This is a major change to the architecture of the current monitor/processor system. There would have to be an new database table to save these "poisonous" jobs and a new system within monitor to deal with them. It may be better to have the processor just take the poisonous job and tack it back onto the end of its own queue to try again later rather than committing suicide. That would be a simpler solution that wouldn't involve making new communication channels with the monitor. Though, I suppose that some jobs could end up repeatedly looping through the queue until the hbase issue was resolved. Do we really want to make a major change to the 1.7 system considering its expected life? Surely any system that resolved the problem for 1.7 would hardly be applicable to 1.8 since that uses a radically different job allocation scheme.
I marked this as a post 1.8 bug. Don't think we should attempt to undertake this for 1.7.x. Even post 1.8, the processors could still run into this type of problem. With the new processors though, we can add an hbase table for storing the exception ooids and implement my strategy in comment #0 fairly easily.
Since processors now recover, this is no longer critical.
Severity: critical → normal
Target Milestone: 1.9 → ---
Component: Socorro → General
Product: Webtools → Socorro
resolved with modern processor
Status: NEW → RESOLVED
Closed: 12 years ago
Resolution: --- → FIXED
You need to log in before you can comment on or make changes to this bug.