Commits on Jan 30, 2014
Commits on Jan 24, 2014
Commits on Jan 23, 2014
  1. Fix TorqueBox.fetch('DLQ') (TORQUE-1189)

    bbrowning committed Jan 23, 2014
    This also fixes TorqueBox.fetch('ExpiryQueue').
Commits on Jan 22, 2014
  1. Let the JBoss clustered session manager handle session expiration (TO…

    bbrowning committed Jan 22, 2014
    Expiring keys behind the clustered session manager's back in
    Infinispan leads to early session timeouts and in certain conditions
    errors upon restarting a clustered TorqueBox.
    This change also removes the default expiration (of less than 2
    minutes) for regular Infinispan cache keys, instead relying on Ruby
    clients to specify a cache expiration when adding entries.
  2. Remove locking from the request path with shared pools (TORQUE-1200)

    bbrowning committed Jan 22, 2014
    This removes several sources of locking that were present in the
    request path with shared pools - locking when borrowing / returning
    Ruby runtimes, locking inside JRuby from initializing RubyIO
    instances, and locking from JRuby when retrieving the JRubyClassLoader.
    With these changes I see about a 50% improvement in top-end
    performance for basic Rack apps. Most applications won't see this
    performance boost, since our locking probably wasn't the bottleneck,
    but for high-throughput and high-concurrency users this should speed
    things up nicely.
Commits on Jan 15, 2014
  1. Only create Sessions in ServletStore if needed (TORQUE-1198)

    bbrowning committed Jan 15, 2014
    Previously we created a new Java HttpSession on every request when our
    ServletStore was used, and now we only create a new one when data gets
    written to the store.
Commits on Jan 14, 2014
Commits on Jan 13, 2014
Commits on Jan 9, 2014
Commits on Dec 17, 2013
Commits on Dec 3, 2013
  1. Fix possible deadlock with bounded pools (TORQUE-1188)

    bbrowning committed Dec 3, 2013
    The zero-downtime deployment work done for TorqueBox 3.0.0 introduced
    the potential for deadlocking when borrowing / returning runtimes to a
    bounded pool. This fixes that issue by splitting a single restart lock
    into separate borrow and return locks so that borrow and return don't
    lock with each other, and instead only lock with the restart action.
Commits on Nov 22, 2013
Commits on Nov 21, 2013
  1. Add "--exclude" option to torquebox archive (TORQUE-1094)

    bbrowning committed Nov 21, 2013
    The --exclude option takes a list of Ruby-compatible regular
    expressions or exact filenames to exclude. The help output of
    'torquebox archive --help' was updated to give an example.
Commits on Nov 20, 2013
  1. remove stray puts

    bbrowning committed Nov 20, 2013
  2. Revert "No more using web CM for polyglot caches, now that I grok the…

    bbrowning committed Nov 20, 2013
    … module attribute"
    This reverts commit b51fb2b.
  3. Revert "Update to newer polyglot with job fix; adjust for mbean insta…

    bbrowning committed Nov 20, 2013
    …ll API changes."
    This reverts commit 99d1964.
  4. Fix reset_session in Rails 4 (TORQUE-1183)

    bbrowning committed Nov 20, 2013
    Prior to Rails 4, Rails took care of clearing the contents of the
    session hash on our behalf after session.destroy was called. In Rails
    4, this was changed to only call session.destroy so we'll clear the
    contents ourselves and invalidate the Java session.
    Without this change, under Rails 4 the user would get a new Java
    session but it contain the exact same session data as their previous
    session, effectively making reset_session do nothing.
Commits on Nov 19, 2013
  1. Revert "Revert rubygems retry hacks to test against chan…

    bbrowning committed Nov 19, 2013
    This reverts commit 8e2839f.
    This is a stopgap until we host our own mirror.