When using a distributed file system like GPFS, a bug in Java 10 can lead to output files from analyses being corrupted. This issue has been observed with CLC software using Java 10. In practice, we have observed the issue affecting analysis log files most frequently. However, if intermediate output files generated during a workflow run are corrupted, this can result in the workflow failing before it completes.
Analyses may complete normally if they are re-launched. This problem, while sporadic, can be frequent enough to be quite inconvenient.
This issue only affects systems running Linux with a distributed file system, such as GPFS. The following software versions are affected:
This issue was fixed in CLC Genomics Server 11.0.1, CLC Genomics Workbench 12.0.1 and CLC Main Workbench 8.1.1 through the introduction of a work-around at the code level that resolves the problem when working on setups where data storage is on GPFS file systems. However, this issue can also affect our internal indexing systems, so we highly recommend that software from affected release lines (CLC Genomics Server 11.x, CLC Genomics Workbench 12.x and CLC Main Workbench 8.1.x) not be installed directly on a GPFS file system.