[S4-44] optional backoff upon multiple consecutive failed checkpoint fetches

• Type: Improvement
• Status: Resolved
• Priority: Major
• Resolution: Fixed
• Affects Version/s: 0.4, 0.5.0
• Fix Version/s: 0.5.0
• Assignee: Matthieu Morel
• Reporter: Matthieu Morel
• Created: 14/Feb/12 01:20
• Updated: 24/Jul/12 16:03
• Resolved: 24/Jul/12 16:03

Description

if a checkpointing backend system becomes unresponsive (e.g. stalled NFS), and that a series of recoveries is to proceed (for instance, startup or failover), then each checkpoint fetching operation will block, wait for a timeout or another kind of exception, and the system will then continue without recovering this PE.

We should provide a way to detect this pattern (multiple backend fetches failures in a short amount of time) and temporarily disable fetching from the backend, in order to reduce blocking when backend becomes unresponsive.

Issue Links

• Is contained by: S4-11 add checkpointing mechanism to s4-piper