JRuby

ObjectProxyCache is not as effective with weakrefs, but eats too much memory with softrefs

Details

  • Type: Bug Bug
  • Status: Closed Closed
  • Priority: Major Major
  • Resolution: Won't Fix
  • Affects Version/s: None
  • Fix Version/s: None
  • Component/s: Java Integration
  • Labels:
    None
  • Number of attachments :
    0

Description

We had to switch the ObjectProxyCache from using weakrefs to softrefs, because softrefs weren't soft enough and the cache ended up eating too much memory. But weakrefs are likely to cause cache entries to be collected before they're ever used, so the new cache isn't helping much.

We need a solution somewhere in the middle so the cache is useful but doesn't bloat to an unreasonable size.

Issue Links

is related to

Bug - A problem which impairs or prevents the functions of the product. JRUBY-1955 Lucene index search causes major memory bloat + slowdown in 1.1RC1, but not 1.1b1

  • Blocker - Blocks development and/or testing work, production could not run
  • Closed - The issue is considered finished, the resolution is correct. Issues which are not closed can be reopened.
relates to

Bug - A problem which impairs or prevents the functions of the product. JRUBY-2904 Need a better solution for tracking Java object wrappers

  • Major - Major loss of function.
  • Closed - The issue is considered finished, the resolution is correct. Issues which are not closed can be reopened.

Activity

Ascending order - Click to sort in descending order
Hide
Permalink
Charles Oliver Nutter added a comment -

This is getting bumped to post 1.1 since it works ok now (maybe not great, but ok) and there's a big Java integration reboot coming.

Show
Charles Oliver Nutter added a comment - This is getting bumped to post 1.1 since it works ok now (maybe not great, but ok) and there's a big Java integration reboot coming.
Hide
Permalink
Charles Oliver Nutter added a comment -

Marking as related to 2904, since a "better solution" would presumably not need to involve weakrefs (or else it wouldn't be much better).

Show
Charles Oliver Nutter added a comment - Marking as related to 2904, since a "better solution" would presumably not need to involve weakrefs (or else it wouldn't be much better).
Hide
Permalink
Charles Oliver Nutter added a comment -

Bill's email from some time ago talking about this problem:

http://jira.codehaus.org/browse/JRUBY-199?focusedCommentId=119902#action_119902

Show
Charles Oliver Nutter added a comment - Bill's email from some time ago talking about this problem: http://jira.codehaus.org/browse/JRUBY-199?focusedCommentId=119902#action_119902
Hide
Permalink
Charles Oliver Nutter added a comment -

I don't see a way that SoftReference will ever work as a mechanism for ObjectProxyCache. It's simply too sticky to use for caching wrappers that may be created by the thousands. WeakReference is not as effective at caching, but we're only talking about lookups at the boundary between Ruby and Java, and if the wrapper is still referenced the WeakReference won't ever deref prematurely.

Marking as WONTFIX.

Show
Charles Oliver Nutter added a comment - I don't see a way that SoftReference will ever work as a mechanism for ObjectProxyCache. It's simply too sticky to use for caching wrappers that may be created by the thousands. WeakReference is not as effective at caching, but we're only talking about lookups at the boundary between Ruby and Java, and if the wrapper is still referenced the WeakReference won't ever deref prematurely. Marking as WONTFIX.

People

  • Assignee:
    Charles Oliver Nutter
    Reporter:
    Charles Oliver Nutter
Vote (0)
Watch (2)

Dates

  • Created:
    Updated:
    Resolved: