Garbage collection issues after 1.2.1 to 1.4.1 upgrade

Hello Elasticsearch community (again)

Following my previous post, that might have been quite unclear, i open this
one because i think i have located the problem (not its source nor its
solution sadly).

So i have an issue with super duper long garbage collectors after upgrading
from 1.2.1 to 1.4.1 . I narrowed the problem down to this looking at the GC
graphs, cluster is running a lot of "full" garbage collectors, taking a lot
(and i mean a lot) of time.
Mmemory used by "old" gc spikes very often ( 5Gb to 32Gb almost instantly )
triggering "stop the world" GC very often.
My memory consumption went from 24 steady gigs to 5, here is the graph
http://we.tl/off4NtOxgN.
Here is a spike : http://we.tl/ADrDyBRHXn (i know right ?, it's a spike) It
happens while there is a "full gc" running.
My java version is "1.7.0_51"

Just so you can see how bad the situation is , see one of the WARN garbage
collector message
[2015-02-28 16:03:48,064][WARN ][monitor.jvm ] [node00]
[gc][old][97830][713] duration [29.6s], collections [2]/[29.6s], total
[29.6s]/[1.8h], memory [5gb]->[4.8gb]/[31.8gb], all_pools {[young]
[74.6mb]->[1.4mb]/[865.3mb]}{[survivor] [0b]->[0b]/[108.1mb]}{[old]
[5gb]->[4.8gb]/[30.9gb]}

I'm neither a sysadmin nor an elastic search expert (and not even a java
dev), so please, ask me for relevant information if you have ANY idea
regarding this problem.

Thank you

--
You received this message because you are subscribed to the Google Groups "elasticsearch" group.
To unsubscribe from this group and stop receiving emails from it, send an email to elasticsearch+unsubscribe@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/elasticsearch/53718314-e8fc-4bf2-80e7-4060775970f4%40googlegroups.com.
For more options, visit https://groups.google.com/d/optout.