resources

March 11, 2015 – Alert: Write Image Journal failure

InterSystems has corrected a defect that may cause data integrity issues following a crash on systems with a certain unusual configuration.  Specifically, the defect can cause application of the WIJ file to fail during startup.

A detailed discussion of the protection provided by the WIJ can be found in the Caché Data Integrity Guide.

This defect is present in all currently released versions of Caché, Ensemble, and HealthShare.  It affects all platforms and operating system versions.

The only systems vulnerable to this defect are those that mount hundreds of databases referenced by long path names.  When the combined length of the path names exceeds approximately 32,000 characters, the defect can occur. The exact condition can be checked for a given system by running the following command:

For Windows:

<install-directory>bincwdimj -d2 -j<wij-directory>

For Unix/Linux

<install-directory>/bin/cwdimj -d2 -j<wij-directory>

For OpenVMS

Define a foreign command

$ cwdimj :== $sys$disk:[-.bin]cwdimj.exe

and from the cachesys manager’s directory issue:

$ cwdimj –d2 –j_<wij-directory>

 

The output of this command will include the value of “Dirlen”.  If Dirlen is less than or equal to 32768, the system is not currently exposed to this defect.  Please note that the value of Dirlen changes as you add or remove databases to your running environment.  If Dirlen is close to the limit of 32768, InterSystems recommends that you request the correction identified below.

If a system exposed to this defect shuts down abnormally, and the abnormal shutdown occurs at a time when there are database blocks pending in the WIJ, startup will fail necessitating manual deletion of the WIJ. This causes database integrity issues. If this occurs, an entry similar to the following will be present in the cconsole log file:

Starting WIJ recovery for ‘<pathname>CACHE.WIJ’.

N>0 blocks pending in this WIJ.

WIJ pass # is 0.

WIJ file does not match expectations: XXXXX

 

The correction for this defect is identified as JPL1675.  It will be included in all future releases of Caché, Ensemble, and HealthShare.  The correction is also available via Ad Hoc distribution from InterSystems Worldwide Response Center (WRC).  If you have any questions regarding this advisory, please contact the Worldwide Response Center.