• Bug
  • Status: Open
  • 2 Major
  • Resolution:
  • cdennis
  • Reporter: jandam
  • July 20, 2011
  • 0
  • Watchers: 1
  • February 17, 2012

Attachments

Description

Stress test tries to flood ehcache with data that can’t be saved to disk store within BACK_OFF_TIME (50ms)

Configuration: LFU, 5 elements in memory, overflowToDisk, eternal

VM args: -ea, -Xmx128M

Test code attached. Please tune parameters DATA_COUNT, ARRAY_SIZE according to you computer speed.

Forum post: http://forums.terracotta.org/forums/posts/list/3993.page#28598

Proposed solutions: *) block putElement till evicted element is stored to disk storage (maybe with timeout) OR *) notify that disk spool is full, let the developer decide what to do.

OOME can be thrown from different part of system that don’t use ehcache

Still reproducible in 2.4.3, 2.5.0b1

Comments

Fiona OShea 2011-07-26

Chris can you take a look at this and see if there is something we can do? Or if there is something we need to do.

Chris Dennis 2011-07-27

This is a real bug that we have been aware of for some time. It’s too late now for this to be fixed in Ulloa, but as you can see this has now been targeted for Vicente which will be the next major Ehcache release.

In my opinion it’s most likely that this will get fixed by using a bounded queue instead of an exception throw however.

Fiona OShea 2011-12-15

Does it make sense to do this in the next Ehcache 2.5.x release?

Chris Dennis 2011-12-19

We’ve not actually had too many complaints about this (really just this one) so I wouldn’t rush it in to 2.5.x. I think the best approach is if I do this early on in Vicente then merge it to 2.5.x when it’s done, it can then just go out with whatever the next 2.5.x release is after that. We’ll also have to explicitly release note for 2.5.x since some people might not like the change in behavior.