Registration

IBM has deployed Samba as part of file storage solutions for many years now. The current product of IBM is called Spectrum Scale and it delivers clustered Samba to a world-wide base of Enterprise customers on top of IBM's clustered file system known as gpfs.

The first part of the talk is supposed to describe selected requirements, use cases and enhancements this mission has been driving over the last years. This could include topics like file contention, access control lists, and Mac support.

The second part of the talk will address the new challenging requirements with respect to identity mapping like the server side group resolution for NFS clients and the increasing demand to add sssd to the clustered Samba server nodes.

IBM Spectrum Scale is a software defined storage offering of a clustered file system bundled together with other services. Samba is included as part of the product to provide a clustered SMB file server and integration into Active Directory. This talk discusses from a development point of view the integration of Samba into a storage product and what the development team has learned over the years. Topics will include the requirement for automated testing on multiple levels and the collaboration with the upstream Samba project. Examples will be used to illustrate problems encountered over time and how they have been solved. Further topics will be challenges that have been solved and gaps that have been seen with the usage of Samba.

In recent times, container application platforms, in particular kubernetes, have become extremely popular, have for instance overtaken the virtual machine centric cloud operating system OpenStack in popularity. In contrast to virtual machine environments which run a variety of operating systems and therefore also have natural use cases for Samba, container platforms are usually running one operating system (kernel) only, and thus don’t offer a very obvious space for Samba, whose main purpose is to act as an agent between different operating systems. Is there still a stage for Samba in container land?

This presentation will start with an introduction to the storage concepts of kubernetes and the container storage interface standard CSI, which generalizes these to other container platforms. It will explain the roles of file, block, and object storage in kubernetes and then shows how a distributed software defined storage system like ceph or gluster is brought into kubernetes, running alongside the consuming applications and managed by so called “operators”, providing storage self service for the applications. Time permitting, the presentation may include a demo of what is possibly the easiest installation of a ceph cluster so far.

From here on, the presentation will explore some very interesting and possibly surprising opportunities for Samba in this environment. Details are omitted in this abstract in order to keep up the suspense.

• Using S4U2Self in winbindd
• The limitations of existing kerberos libraries
• The challenges of adding new features to kerberos libraries
• Kerberos testing with plain python

Testing is an extremely important part of the software development process. Luckily, a lot of the testing work is automated today. Continuous integration (CI) is a buzzword. In Samba, we have come a long way from manually running tests (sometimes) to our own autobuild system enforcing a full run of the test-suite as a push-gate to running various tests in parallel in the gitlab ci system for each merge request.

While testing server client systems like Samba in an automated way is demanding already, testing a cluster is even more complicated and resource hungry. CTDB itself has been tested in isolation with local processes since the beginning, and recently a test-environment has been added to Samba’s selftest that helps test the samba+ctdb stack entirely with local process and socket wrapper. But an automated, periodic test of a samba and ctdb setup on top of a real clustered file system is still missing. One aspect why this is not so easy to implement is the fact that it would usually require a couple of virtual machines to set up such a test cluster.

This presentation will introduce a project that we recently started to investigate and work on. It aims to create a periodic test run pulling the latest bits of Samba and Gluster, setting up a cluster and running test suites against it. Errors would be reported to both projects. As compute resources for the test runs, the centos-ci is used. This project provides jenkins-managed bare metal server resources for open source projects to integrate into their CI systems. These servers are powerful enough to run realistic cluster setups in virtual machines. The presentation will demonstrate how the centos-ci resources are integrated into this test system. Furthermore, possibilities will be explored, how to integrate centos-ci resources as additional runners for Samba’s gitlab CI runners.

From the network down to the filesystem and back. This talk will give an overview on Samba's SMB2 packet processing with the goal of giving novice smbd hackers a starting point to the most important source code subsystems involved when a client sends a "create a file" request to the server.

This is a report on the status of CTDB, similar to that presented at recent SambaXP conferences. The report will focus on 2 main areas: progress and plans. What is new upstream in CTDB? What do our plans look like compared to those presented in recent years? There is also the intersection of progress and plans: what useful things are sitting in development branches but are not merged?

Progress includes: Clustered Samba testing is now in Samba's test suite and autobuild, CTDB's inter-node TCP transport is now more resilient (with some pain along the way), database vacuuming has been simplified, the recovery lock has been enhanced, code is generally cleaner (largely due to csbuild showing issues) and there have been many improvements in testing.

Plans include: splitting CTDB into multiple daemons (as previously presented), a transport using datagram messaging and simple code for new developers to understand and embrace.

Recently spotlight got support for elasticsearch, what about the experimental WSP support?
1. Can it do the same?
2. What's involved
3. How does it affect the existing implementation

This is a follow up to last years talk about managing AD via an ncurses gui. This talk will cover additional improvements to the ADUC and ADSI Edit modules, as well as covering a new DNS Manager. ADUC and ADSI Edit now communicate with AD via samba python bindings, and the DNS Manager interacts with samba-tool calls. The tools have also been wrapped in a redistributable AppImage, which can run on multiple distros. Automated testing has also been written.