Cinder

Ceph driver should have backoff_rate retry option

Bug #1798346 reported by Pavel Abalikhin
14
This bug affects 2 people
Affects Status Importance Assigned to Milestone
Cinder
Incomplete
Undecided
Unassigned

Bug Description

Nowdays RBD driver supports rados_connection_retries and rados_connection_interval options. But utils.retry also have backoff_rate parameter and it would be nice to have ability to set it.

Tags: ceph drivers rbd
Revision history for this message
Sean McGinnis (sean-mcginnis) wrote :

What bug does this cause not having this a configurable option?

Not sure we would want to expose that, but if you explain the bug you are encountering we can see if it would be something necessary.

Changed in cinder:
status: New → Incomplete
tags: added: ceph drivers rbd
Revision history for this message
Pavel Abalikhin (anpavl) wrote :

In the case when volume can't be deleted immediately (active flattening for example) retry tries to do it again and again but always using backoff_rate = 2 that means that interval between attempts is growing. It can be inconvenient for Nova when it tries to clean up instances which are erroneously still running after having been deleted or for cloud operators who want to change the default behaviour. So actually it's not a bug but an improvement.

To post a comment you must log in.
This report contains Public information  
Everyone can see this information.
Subscribing...

Other bug subscribers

Remote bug watches

Bug watches keep track of this bug in other bug trackers.