internal service calls cannot handle interrupted requests
Bug #740674 reported by
Robert Collins
This bug affects 2 people
Affects | Status | Importance | Assigned to | Milestone | |
---|---|---|---|---|---|
Launchpad itself |
Triaged
|
Low
|
Unassigned |
Bug Description
When an internal service fails its load balancer (haproxy today, lvs+haproxy in the future) cannot resurrect already dispatched requests. We will see a cluster of oopses at the time that that occurs - those oopses are genuine and we should see them (we may need to take some remedial action, for instance). Where its safe, we may want to automate the remedial action (which for idempotent services (like GET requests)) could well be to re-request and not-oops about the initial error.
We expect a vanishingly small error rate here - something like 1 failure per 20 million front end requests)
tags: | added: branch-puller |
description: | updated |
Changed in launchpad: | |
assignee: | nobody → Aaron Bentley (abentley) |
summary: |
- ConnectionRefusedError from code puller + Code puller can't access XMLRPC server |
description: | updated |
Changed in launchpad: | |
assignee: | Aaron Bentley (abentley) → nobody |
description: | updated |
To post a comment you must log in.
I believe branches have a field for showing connection errors already;
there's no particular need to log this if the data is put into that
field.