• Print |
• E-mail  |
• Bookmark

If a cluster has multiple Object Spawners associated to it (ie. the
cluster spans different machines), some or all of the Object Spawners
involved may become deadlocked.  This deadlock can occur when the Object
Spawners involved are started simultaneously (such as when started by a
common shell script or batch job).

The problem occurs when two (or more) Object Spawners connect to each
other at the same time and start communicating with each other at the
same instant.

During the startup and initialization of the Object Spawners in a
cluster, Object Spawners that are members of the same cluster must
initialize communication for load balancing to work across them.  The
initialization process includes binding to the remote Object Spawner's
port and obtaining an interface to the remote Object Spawner's load
balance.  The Object Spawner then sends a copy of it's own interface to
the remote Object Spawner.

This results in two (or more) Object Spawners believing they are at the
top of the hierarchy and try to update the same data at the same time.
A deadlock occurs when the Object Spawners are going after the same
resources.  This causes the Object Spawners involved to be unable to
redirect any clients.

A workaround for 9.1.3 is to have a "sleep time" in between each Object
Spawner startup.

A fix for SAS 9.1.3 (9.1 TS1M3) for this issue is available at:

http://www.sas.com/techsup/download/hotfix/e9_sbcs_prod_list.html#014276

For customers running SAS with Asian Language Support (DBCS), this
fix should be downloaded from:

http://www.sas.com/techsup/download/hotfix/e9_dbcs_prod_list.html#014276

Operating System and Release Information

A fix for SAS 9.1.3 (9.1 TS1M3) for this issue is available at:

A fix for SAS 9.1.3 (9.1 TS1M3) with Asian Language Support (DBCS) for this issue is available at :