July 27, 2017 – Alert: Linux Defects Can Corrupt Mirror Copies of Journal Files

July 27, 2017 – Alert: Linux Defects Can Corrupt Mirror Copies of Journal Files

InterSystems has encountered defects in Linux which can corrupt copies of journal files that are generated on a mirror backup or async member; this occurs only in certain specific configurations. The original mirror journal file created on the primary member is not affected.

The risk is only exposed in Caché, Ensemble, and HealthShare distributions beginning with version 2017.1.0. The risk only exists on Linux and only if a backup or async mirror member is configured to use the rtkaio library.

The rtkaio library is distributed with Red Hat Enterprise Linux, though it may be possible to acquire it for other Linux distributions. Caché does not use the rtkaio library by default. Therefore, the risk does not exist unless your system has been configured to use rtkaio. Sites most commonly configure Caché to use rtkaio via the LibPath parameter in cache.cpf.

When the problem is triggered, copies of mirror journal files retrieved on a backup or async mirror member may become corrupted. This typically results in errors during dejournaling of the corrupt file and an alert in the cconsole.log similar to

03/29/17-16:18:12:852 (25129) 2 Dejournal stuck at 7342380 (0x0070092c) cyclejrnend=11141120 in file #9 

InterSystems has notified Red Hat about this issue. Until a correction becomes widely available in supported Linux releases, InterSystems offers a workaround for sites that have configured Caché to use rtkaio. This workaround automatically removes any occurrence of rtkaio from the library path for the daemon process that writes these copies of mirror journal files.

Note that the workaround is not effective if rtkaio use was configured by manipulating the library files rather than setting the LibPath parameter.

The workaround is identified as RJF264 and will be included in all future releases of Caché, Ensemble, and HealthShare. The workaround is also available via Ad Hoc distribution from InterSystems Worldwide Response Center (WRC).

If you have any questions regarding this alert, please contact the Worldwide Response Center.