too_happy with MultipleHosts() after all but one hosts are full

[GoogleCodeExporter]

What steps will reproduce the problem?
1. get two storage nodes
2. fill one to 100%, second must have >= 2 devices
3. files replicate to 2 devices on non-full host
4. add 3rd host or free some space on full host
5. too_happy: fid ... is on too many devices. It just goes on like this.
6. fsck seems to do nothing about this

What is the expected output? What do you see instead?
i expected it to delete the extra file on same host and replicate it to another 
host after extra space is available

What version of the product are you using? On what operating system?
p5-MogileFS-Server-2.57
FreeBSD 8.2

Please provide any additional information below.

Original issue reported on code.google.com by Gle...@gmail.com on 21 Mar 2012 at 7:21

[GoogleCodeExporter]

Any chance you could attach the `mogfiledebug` output from one of those fids 
which has been printed in the "too_happy" log notices?

Replication should retry a few times per fid and end up fixing itself. 
mogfiledebug will tell us how long a fid's been bouncing around there.

Original comment by dorma...@rydia.net on 21 Mar 2012 at 11:18

[GoogleCodeExporter]

Sorry, I already deleted & reuploaded them.
It was about 10k fids with 7+ failcount in file_to_replicate.
Waited just for a few hours after i added space and file_to_replicate didn't 
start descreasing - had no idea it could start trying to do something else 
later. No new files were added since adding new space.

Original comment by Gle...@gmail.com on 22 Mar 2012 at 4:28

[GoogleCodeExporter]

We have a few of these over replication bugs still. I'm going to schedule time 
for myself in july to ensure they're taken care of, if someone else doesn't get 
them first.

Original comment by dorma...@rydia.net on 20 Jun 2012 at 1:00

• Changed state: Accepted