object auditors don't finish
Affects | Status | Importance | Assigned to | Milestone | |
---|---|---|---|---|---|
OpenStack Object Storage (swift) |
Fix Released
|
Critical
|
Christian Schwede |
Bug Description
I mean that they don't finish often, not at scale, at least not often enough that you're likely to get a 'completed' log line before you release new code or change a config and restart them.
I *think* the best thing would be write out the device you're on when you start, and pick up from there when you restart. I'd appreciate some other deployers who are at scale chiming in if they think a dev is a good unit of work to potentially restart. We could go to partition?
I'd probably want to log a big happy message when you finish a device (or whatever unit of work), and then another one when you roll over from sdzd to sda. From there you should be able to start to monitor your auditor cycle times and decide if they're within a reasonable threshold.
I think it could be fixed once in swift.common.
What happened with bug # 691884
description: | updated |
Changed in swift: | |
status: | Expired → Confirmed |
Changed in swift: | |
importance: | Undecided → High |
Changed in swift: | |
assignee: | nobody → Christian Schwede (cschwede) |
Changed in swift: | |
importance: | High → Critical |
Clay, what do you think of this 2-year-old bug report?