Shoal 1.1 bits on the Shoal download site have been updated to reflect
the promoted build dated 10/21/2008.
Get the promoted build *here
<https://shoal.dev.java.net/downloadsindex.html>
*
We have also added a buildable Source and Javadocs zip distribution
corresponding to this promoted build on the download page based on a
community user request.
Changes that went into this build after the previously promoted build
(08202008) are :
* Fix for sailfin issue 1276
<https://sailfin.dev.java.net/issues/show_bug.cgi?id=1276>,
non-timed wait in getMemberStateViaLWR. Confirmed by running
developer level test that issue was fixed. Also confirmed on SUSE
Linux test setup of 10 instance Sailfin cluster
* Fix for sailfin issue 1119
<https://sailfin.dev.java.net/issues/show_bug.cgi?id=1119> . Made
call to getMemberState() reentrant. Documentation on how to use
this API will be added to the issue.
* Fix for Sailfin issue 484
<https://sailfin.dev.java.net/issues/show_bug.cgi?id=484> With the
fix, when the master is killed, a new master is elected and that
master sends out a Join and JoinedAndReady notification for the
previously killed master.
* Sailfin Issue 1233
<https://sailfin.dev.java.net/issues/show_bug.cgi?id=1233> :
Logging issue
* 1197 <https://sailfin.dev.java.net/issues/show_bug.cgi?id=1197> :
Partial fix for memory leak issue portion coming from
HealthMonitor that caused connection checking on multihomed IPv6
enabled connections to hang while spawning multiple threads.
* 1171
<https://sailfin.dev.java.net/issues/show_bug.cgi?id=1171>:[blocking]
Memory leak when SSR is enabled for st1_3_iptv_reinvite scenario
solution was to incorporate fix from Sailfin Issue 1128 (see below)
* 1128 <https://sailfin.dev.java.net/issues/show_bug.cgi?id=1128>:
Major leak in idle systems
* 1119 <https://sailfin.dev.java.net/issues/show_bug.cgi?id=1119>
Partial fix. Need API to get cluster members with their states.
Added a new api that allows clients to query member state based on
freshness of the last heartbeat within a threshold time and for
those beyond the threshold make a network call to the concerned
member about its health state under a specified timeout.
* Partial fix for Shoal issue : 79
<https://shoal.dev.java.net/issues/show_bug.cgi?id=79>:
DistributedStateCacheImpl not thread safe.
* Number of other fixes
<https://shoal.dev.java.net/servlets/ReadMsg?list=cvs&msgNo=667>
that did not have a bug id but based on analysis of test runs from
project Sailfin's tests.
* Remove unchecked cast compiler warnings. Make
ClusterViewManager.masterViewID thread-safe.
* Fix for sailfin issue 626
<https://sailfin.dev.java.net/issues/show_bug.cgi?id=626> in
ClusterManager.getFailureDetectionTcpRetransmitTimeout. Change
default timeout from 30 seconds to 10 seconds. Total time it takes
to detect that a machine with a pulled network cable is now
approximately 30 seconds. Default glassfish heartbeat is 2 seconds
with 3 retries. There is a wait of 2 seconds before failure
verification. There are 2 calls to isConnected() with timeouts of
10 seconds(fixed this with this change).
* Issue 1006
<https://sailfin.dev.java.net/issues/show_bug.cgi?id=1006> : Fix
for in isConnected() and class CheckConnectionToPeerMachine
designed to help Shoal detect hardware type failures within a
threshold.
* Fix for glassfish issue 5313
<https://glassfish.dev.java.net/issues/show_bug.cgi?id=5313>:
enable specifying which network interface for Shoal to use on
multihome machines by setting gms-bind-interface-address.
Documentation on how to set this configured is in glassfish issue
5313.
* Fix for Shoal issue 45
<https://shoal.dev.java.net/issues/show_bug.cgi?id=45> :
MessageWindow NullPointerException handling GMSMessage
Thanks very much to Sheetal, Joseph Fialli, Mohamed Abdelaziz, Bongjae
Chang, Sailfin and GlassFish Quality Engineers Kazem, Sony and Varun,
and most of all the Shoal community members.