December was largely dominated by the holidays, yet we have managed to close a few issues for our upcoming 4.1.8 release in support of Membership migration. We have also continued work on the 4.2 rule engine plugin framework within the experimental-build-system branch.
Federation testing is nearly complete as is the migration of the rest of the plugins. We still need to expand platform coverage on the Kerberos and GSI plugins. Once that is complete Run-In-Place will be next on the list for automated testing.
We have been working closely with one of our members during their migration from the legacy 3.3.x system to the latest in the 4.1.x series. As issues have been identified we have slated them for a near-term 4.1.8 release. We have also included several issues from the community and others that have already been fixed for 4.2.
Of note, we have added a capability to the unix file system plugin: a high water mark. This context string option will allow a user to set a high water mark for the file system which will report “Out of Free Space” to iRODS if adding a new replica will push the used space past the number of bytes specified. Coupled with a Round Robin or Random coordinating resource, this will effectively create Read Only child resources while allowing head room for additional writes to existing replicas. The parent resources will simply try another child if the current child votes “no” for a create operation.
Additionally, we have significantly improved the reliability of iRODS under heavy network load, as well as improved the interaction of some system microservices with resource hierarchies.
The majority of the work for 4.2 has been integrating the rule engine plugin framework. The code was merged and then brought up to date with the newly implemented C++14 plugin architecture. This new code is now passing all existing tests. The next step will be to add Avro serialization of all plugin operation parameters which will be passed to the dynamic policy enforcement point. This new serialization will normalize the behavior across all possible rule engine plugins while giving the policy author maximal information when making policy decisions.