How does reducing replicas work?

asked 2016-06-09 13:44:32 -0600

Nelson Marcos gravatar image

Hi everyone!

Our Swit cluster has two region and we were using 4 replicas. Region 2 has more devices.

We decided to reduce it to 3 replicas. We setup the ring and deployed it. Everything went fine. But a after few hours, disk usage on Region 1 started to increase and disk usage on Region 2 started to decrease. After 24 hours Region 1 disk usage started to decrease and become very close do Region 2.

(http://imgur.com/looXsTA)

Can someone explain why this behaviors happened? Does it have any relation with min_part_hours (24 hours) ? Even if it has, why would it increase on one region instead of just maintain the same usage?

I thought, after the reduction, swift would just remove the extra copies.

PS: we had not outage and nobody complained about anything. I just want to understand a little bit more how swift works.

edit retag flag offensive close merge delete

1 answer

Sort by ยป oldest newest most voted
0

answered 2016-07-20 17:16:36 -0600

updated 2016-07-20 17:19:14 -0600

Best resource to understand what happened is this document under Dispersion and Overload.

edit flag offensive delete link more

Get to know Ask OpenStack

Resources for moderators

Question Tools

1 follower

Stats

Asked: 2016-06-09 13:44:32 -0600

Seen: 123 times

Last updated: Jul 20 '16