Repeated tranlog space full condition can cause fatal stone error
The stone process is designed to pause when tranlog disk space is
exhausted, giving the system operator a chance to provide space by
either deleting files or adding additional tranlog directories
(see section 6.11 of the GemStone/S Systems Adminstration Guide for
details). When space has been cleared, the stone will then continue.
But on Solaris, if tranlog disk space is exhausted/cleared several
times, there is a risk that the stone will get a fatal system error.
The type of system error depends on the version of GemStone:
GS/S 5.x: SIGBUS Bus Error on invalid address alignment
GS/S 6.x: SIGSEGV Segmentation Violation on errorno=78 ENAMETOOLONG
Workaround
Make sure that you never run out of disk space on directories holding
tranlogs. All production systems should have a background task that
monitors available extent and tranlog disk space and warns
the operator when space drops below some threshold.
Last updated: