Services
& Support

Support Alerts 2012


October 30, 2012 – Advisory: CSP Gateway and Apache on AIX

InterSystems has identified several issues that can lead to the CSP Gateway becoming unresponsive and cause applications to timeout when using Apache HTTP Server on AIX.

These issues only occur on configurations using Apache HTTP Server running on IBM AIX and affect Caché, Ensemble and HealthShare.

Please note that InterSystems Management Portal uses a private server installation of Apache therefore it is also susceptible to timeouts.

There are three discrete issues.

  • Issue #1
    The Apache parent process dies with Signal 11 errors evident in the Apache server error log file.  These process failures are caused by a bug in Apache’s implementation specific to AIX OS.
  • Issue #2
    Using the SHM=OS setting triggers a memory leak in Apache.
  • Issue #3
    This is a defect in InterSystems CSP Gateway; Processes are not fully released and are indefinitely in the “closing” state in the CSP Gateway process table.

InterSystems believes that the first two are related to the interaction of Apache and AIX.  With regards to these issues we recommend the workaround documented below.  The third issue is a defect in InterSystems CSP Gateway component – a correction is available.

The fatal errors caused by Issue #1 can be avoided by setting the parameter SHM=OS in the CSP.ini file. It is necessary to restart the CSP Gateway for this change to take effect.

To completely avoid both issue #1 and #2, set SHM=OS and employ the NSD (Network Service Demon) approach for the CSP Gateway rather than the managed modules that are installed by default with Apache.  Note: NSD is not generally InterSystems preferred approach and is recommended only as a workaround to these known issues.

Once we believe these issues have been resolved, we will provide a further advisory with an updated recommendation.

The correction for Issue #3 is identified as CMT1038.  It is included in Caché and Ensemble as of versions 2012.1.2 and 2012.2.0.  It is included in HealthShare as of version 2012.1.0.  It is also available by request as an Ad Hoc distribution for earlier versions.

An alternative approach to avoid the risk of all three issues is to deploy Apache on a Windows or Linux server, or to deploy IIS.

If you have any questions regarding this advisory, please contact InterSystems Worldwide Response Center (WRC).


September 6, 2012 – DeepSee Alert: Potential cell cache problem

InterSystems has corrected a defect that could cause DeepSee MDX queries to produce incorrect results.

This defect is present in all releases of Caché and Ensemble prior to 2012.2 build 638 and on all platforms.

DeepSee cubes with more than 512,000 facts are at risk; this covers the majority of customers.  Please note that for cubes using a bucket size less than the default of 8 the problem can occur in cubes with fewer than 512,000 facts.

A correction for this defect is available and is identified as JMD1115.

This correction will be included in the released version of 2012.2 and all subsequent releases.  The correction is also available via Ad Hoc distribution.  InterSystems recommends that all systems at risk apply this correction, either by upgrading to the upcoming release of 2012.2 or by requesting an Ad Hoc distribution.

If you have any questions regarding this or to request an Ad Hoc, please contact InterSystems Worldwide Response Center (WRC).


September 5, 2012 – Alert: Potential System Hang and Data Integrity Problem

InterSystems has corrected a defect that may result in hung systems and data corruption in environments using InterSystems database encryption on AIX and 64-bit Red Hat Linux.

This defect is present in all releases of Caché and Ensemble from 2010.2.1 to 2012.1.0 (inclusive).  Platforms running any version of AIX are at risk for this. Platforms running 64-bit Red Hat Linux where the version string contains “Red Hat Enterprise Linux 5 for x86-64” are at risk.

There are two failure outcomes for this defect. The first results in a hung system and potentially corrupt databases on a subsequent restart. The second involves potential database corruption upon restarting after any system crash (for example, a power outage or ccontrol force).

These failures only happen on systems with 8 or more cores that use InterSystems database encryption.  In order for database corruption to occur Write Image Journal (WIJ) blocks must have been restored during the restart. WIJ block restoration occurred if the following message appears in the cconosle.log: ‘Operator: “Is an operator present to answer questions (Y/N)? …’.

If you encounter a hang on a vulnerable system, InterSystems recommends contacting the Worldwide Response Center for help resolving the problem and avoiding data corruption.  DO NOT force your system down first!

If you suspect you may have encountered this problem, InterSystems recommends that you run an integrity check of all databases.

A correction for this defect is available and is identified as WDS426; it is included in Caché 2012.1.1 and will be included in all future releases including maintenance releases. InterSystems recommends that installations at risk request an Ad Hoc distribution to avoid exposure to this defect.


May 11, 2012 – Alert: Database Compaction and Database Defragmentation May Cause Corruption

InterSystems has identified a defect in the Database Compaction and Database Defragmentation feature that can cause the target database to become corrupt.

The Database Compaction and Database Defragmentation features were introduced in Caché and Ensemble versions 2010.2, and all releases 2010.2.0 and later are at risk.  This impacts all Operating Systems, except VMS where these features aren’t currently enabled.

The older global compaction utility, ^GCOMPACT, is not affected.

InterSystems recommends discontinuing the use of Database Defragmentation and Database Compaction.  An additional alert will be posted when a correction is available.

If either of these features, which appear in the ^DATABASE utility as options 13 (“Compact freespace in a database”) and 14 (“Defragment globals in a database”), have been used, an integrity check should be run against the database(s) in question to detect if any problems were introduced.

In addition, InterSystems is removing these options from the ^DATABASE menu for 2012.2, and for any future maintenance release beyond 2012.1.2.

If you have any questions regarding this, please contact InterSystems Worldwide Response Center (WRC).


April 30, 2012 – Advisory: Network Interruptions on AIX-based Systems

InterSystems has received reports of slow TCP data transfer on IBM AIX-based machines, leading to network interruptions and TCP timeouts.  Application of corrections from IBM has been found to address these reports.

If you have experienced similar problems you may want to consider these corrections.

If you have further questions regarding this advisory, please contact the InterSystems Worldwide Response Center (WRC).


April 26, 2012 – Update – Alert: Database Defragmentation or Database Compaction May Cause Corruption

InterSystems has corrected the defect outlined in the alert: “January 13, 2012 –Alert: Database Defragmentation or Database Compaction May Cause Corruption”

In this earlier alert InterSystems recommended discontinuing the use of the Database Defragmentation and Compaction. With the corrections announced here, it is now safe to use these functionalities.

The relevant corrections are identified as: JO2499, JO2503, JO2504, JO2508, SAP1374, SAP1382.

These corrections are included natively as of the following distributions of Caché and Ensemble:

  • 2010.2.8
  • 2011.1.4
  • 2012.1.0

The corrections are also available via Ad Hoc distribution from the WRC.

If you have any questions regarding this, please contact InterSystems Worldwide Response Center (WRC).


January 13, 2012 – Alert: Database Defragmentation or Database Compaction May Cause Corruption

InterSystems has identified a defect with Database Defragmentation and Database Compaction that can cause corruption of a database. Note: the Legacy GCOMPACT utility is not affected by this defect.

This defect affects all currently released Caché and Ensemble versions beginning with 2010.2, on all platforms and operating systems except VMS (the Database Defragmentation and Database Compaction features are not available on the VMS platform).

InterSystems recommends discontinuing the use of Database Defragmentation and Database Compaction.  An additional alert will be posted when a correction is available.

If either of these features has been used an integrity check should be run against the database(s) in question to detect if any problems were introduced.

If you have any questions regarding this, please contact InterSystems Worldwide Response Center (WRC).