Description: Affected Components for 5.1.3_hotfix08
The following components are updated by this hotfix:
Resin
Manager
Engine
Windows Daemon
.NET Services
================================================================================
Closed Issues in 5.1.3_hotfix08 (This Release)
C++ Specific
GS-12889
In heavily multi-threaded applications, it was possible for the Driver to
deadlock when performing Service cancellations.
GS-12682
It was possible for the Driver to crash when submitting a large amount of
services from multiple threads.
GS-12982
A failure to decrypt data would block the cpp Driver and floods the broker logs
with error messages while retrying. We have now implement a temporary workaround
where the cpp drive will throw a LiveclusterException in order to allow the
client code to take control of the resubmission logic. We are still working on a
long term solution that does not require any user intervention.
GS-12997
Driver now report some extra diagnostic information at startup such as platform
and compiler.
GS-12998
We have added extra logging at FINE level to display submission information for
the cpp driver.
GS-13002
When using compression on a cpp Driver, large payloads would cause the Driver to
crash.
GS-13005
We have added the ability to run callback threads in parallel for cpp Drivers.
In order to enable this, the environment variable DS_CALLBACK_PARALLEL needs to
be enabled and set to "true".
Java Specific
GS-12965
Upon a disconnection, it was possible for a Java Driver to be stuck in a loop
for an indeterminate period of time while trying to reconnect. This would
manifest itself in the logs as a failure to connect due to an XML Exception with
the following message: " in bean MessageWrapper property message, may be missing
set or get."
Stability,Fault Tolerance
GS-12767
When the Manager was sending notification messages to the Daemon, such as Task
assigned or cancelled, the Daemon's fileserver was instructing the Manager to
retry after 5s if the Daemon was not logged in to a Director. This was
unnecessary and could cause problems with backlogging of notification messages.
(Note that this situation was addressed in a prior hotfix by having the Manager
ignore the retry directive.)