Caché News, Alerts, and Advisories

Receive these Caché News, Alerts, and Advisories by E-mail

May 3 , 2007 –Possible Caché Data Corruption with ECP

InterSystems has corrected a Caché defect that under rare conditions can lead to missing and corrupt data in ECP (Enterprise Cache Protocol) environments.

This defect exists in all currently released Caché versions that support ECP and for all platforms and operating systems.

The necessary condition to trigger this defect is a process running on an ECP client referencing data on multiple ECP servers.

Please note that even when this condition exists in an environment the likelihood of encountering this defect is low.

The correction for this defect, identified as GK589, is included in the upcoming Caché 2007.1 release. It is also available from InterSystems in an Ad Hoc dis

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

April 19, 2007 - Caché and Ensemble on Microsoft Vista

InterSystems has identified a number of characteristics of Microsoft Vista that require changes to Caché and Ensemble.

Starting with Caché 2007.2 and Ensemble 2007.2, InterSystems will fully support Microsoft Vista Business and Microsoft Vista Ultimate as a server platform, and all editions for Microsoft Vista as a client platform.

While Microsoft Vista is not supported for production environments with versions prior to 2007.2, InterSystems will assist customers who want to experiment with Vista and Caché 2007.1.  We have assembled a list of currently-known workarounds to run Caché 2007.1 on this operating system version.

Please check http://www.intersystems.com/vista for updates to this document as we become aware of further Microsoft Vista related issues.

For additional information, you can contact InterSystems through the WRC Online system or by contacting InterSystems Support directly at support@intersystems.com.

Caché 2007.1 is available for customers as part the Developer Download Program.

Documentation

 

April 11, 2007 – Advisory: Caché and Ensemble on 64-bit Microsoft Windows

InterSystems has discovered an issue with Windows memory management that has important implications with the larger Caché buffer pools possible on 64-bit Windows platforms. The issue is that memory requirements on Windows platforms are larger than on similarly configured UNIX Platforms.

Large buffer pools (i.e. large shared memory segments) require that each Caché or Ensemble process set aside a significant amount of private memory for internal use by Windows (called Page Table Entries or PTEs). On 64-bit Windows, where very large buffer pools can be defined, the cumulative effect can be considerable. To ensure good performance sufficient physical memory must be available - paging of the PTEs can lead to very poor performance,

For example, a 16GB shared memory section will require 32MB per process of Windows internal memory management tables. Thus 1000 processes will require 32GB of memory just for page table entries.

InterSystems can assist with calculating memory requirements for customers considering a 64-bit Windows platform.

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

March 11, 2007 – InterSystems Products and VMware Virtualization Software

InterSystems has determined in benchmark and customer testing that VMware environments running Caché and Ensemble require significantly greater CPU capacity than native installations on the same hardware.  In multi-user tests, CPU load is typically twice as great (200%) of the same tests run on the same operating system natively (i.e. without VMware) on the same hardware.  Additionally, there are application specific implications for response time for interactive jobs.

InterSystems, along with a number of its customers, have reported this problem to VMware and VMware is investigating.  For updates on the status, please contact VMware Global Alliance team.

This affects all versions of all InterSystems products.  Until a solution is available from VMware, InterSystems recommends that any customers contemplating a VMware deployment

  • Ensure that available CPU capacity is at least twice as large as a native installation and
  • Conduct application specific testing to ascertain what, if any, response time impact will be noticed by users.

Please contact InterSystems Worldwide Response Center (WRC)  before deploying Caché or Ensemble on VMware.

March 6, 2007 – Caché, Ensemble and Microsoft Vista

InterSystems testing has revealed a number of issues that prevent installation and operation of existing Caché and Ensemble versions on Microsoft Vista.  Many of these are related to new operating system security features.  Based on what we have discovered to date -- and we need to keep in mind that more issues may yet surface -- our plan is as follows:

  1. The 2007.2 versions of Caché and Ensemble will fully support Microsoft Vista for both server and client configurations.
  2. It is possible that the 2007.1 versions of Caché and Ensemble will run on Vista with certain limitations and some manual "work arounds".  For instance, Vista introduces a new version of Microsoft's IIS Web server which drops support for the mechanism (called ISAPI) which is used to connect CSP to IIS.  The work around is to manually re-enable ISAPI support in IIS.  (Alternatively, customers may decide to use Apache instead of IIS.)  In 2007.2 we will move to a new IIS connection mechanism for CSP.
  3. Caché and Ensemble versions prior to 2007.1 will not be supported on Microsoft Vista.

From a licensing standpoint, Microsoft Vista is not a new platform.  We will continue to have three Windows platforms: x86 32 bit, x86 64 bit, and Itanium.  Moving from one Windows version to another does not constitute a platform change.  Moving from one processor family to another is a platform change, whether or not the Windows version changes.  For instance:

  • Moving from Windows XP on an x86-64 computer to Vista on that same computer is not a platform change.
  • Moving from Windows XP on an x86-32 computer to Vista on an x86-64 computer is a platform change.  (Making the same change from an x86-32 computer to an x86-64 computer with Windows XP on both is also a platform change.)

If you have any additional questions about our plans for Vista, please contact Andreas Dieckow (Andreas.Dieckow@intersystems.com).

February 11, 2007 - Journal Start Failure in Caché Cluster

InterSystems Worldwide Response Center (WRC) has identified a defect which in rare cases could lead to journaling failing to start in a Caché cluster.

This defect is present in all currently released versions beginning with 5.1. Only Caché concurrent cluster configurations on OpenVMS or Tru64 are at risk.

The problem can only occur when the Caché cluster master instance is shut down normally after another instance in the cluster is shutdown or crashes. The result of this defect is that an invalid journal file name may be recorded in the write image journal of the Cache cluster master instance that was shut down. If this occurs, then journaling may fail to start upon subsequent startup of that instance of Caché. Even with the necessary conditions met, the problem occurs rarely and unpredictably.

The correction to this problem is identified as GK579 and will be included in the final release of Caché 2007.1. It is also available from the WRC in an Ad Hoc distribution.

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

February 8, 2007 - Journal Corruption with Caché Clusters as an ECP Database Server

InterSystems Worldwide Response Center (WRC) has identified a defect which can cause journal file corruption in a Caché cluster when the Caché cluster acts as an ECP server to external ECP clients. 

This defect is present in all currently released versions beginning with 5.1.  Only Caché concurrent cluster configurations on OpenVMS or Tru64 are at risk.

The corruption, if triggered, will occur during cluster failover when the Caché cluster master shuts down or crashes while ECP clients are connected with open transactions on the cluster.  The corruption will occur in the new master’s current journal file.   As a result, transactions requested to roll back may not roll back properly, and journal files may not be usable for disaster recovery.  Additionally, any shadow destinations connected to the cluster as the source of shadowing may fail or have incorrect data.

The correction to this problem is identified as HYY1291 and will be included in the final release of Caché 2007.1.  It is also available in an Ad Hoc distribution.

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

January 5, 2007 – Shadow Data Corruption

InterSystems Worldwide Response Center (WRC) has identified a defect which could lead to missing or incorrect data on the destination of Caché shadowing.

This defect is present in all currently released versions beginning with 5.0.13. This includes 5.0.13 - 5.0.21, 5.1 and 5.2 – 5.2.3. All platforms and operating systems are at risk. There is no risk to data on the shadowing source system, only data on the destination systems is at risk.

The defect is triggered only (but not always) when shadowing is suspended or stopped, or when the destination of shadowing is shutdown. The problem may occur without any reported error. The missing data will not be reapplied when shadowing is restarted.

A configuration change will protect against this problem. Since this configuration change may slow shadowing performance considerably it is recommended only as a temporary measure until the correction can be installed. To implement the protection:

  • On the destination of shadowing, in the %SYS namespace, set global node ^SYS("shdwcli",sys,"func")="DejrnCOS" where 'sys' is the name of shadow.

  • Suspend and resume shadowing (stop/start in 5.0.x)

The correction to this defect is identified as HYY1280 and will be included in the final release of Caché 2007.1. It can also be requested in an Ad Hoc distribution. The correction addresses the problem for fast mode shadowing, which is the only type of shadowing available in Caché 5.1 and above. In Caché 5.0 two types of shadowing existed: fast mode and compatible mode. Instances using compatible mode are recommended to implement the configuration change protection documented above as a permanent solution.

Note: It is only necessary to apply the correction to Caché instances that act as a destination ofCachéshadowing.

IMPORTANT: After applying the correction or configuration change it is recommended to resynchronize the destination of shadowing with a copy of the databases from the source in order to ensure that there are no data integrity problems.

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