- 1 Introduction
- 2 Rationale
- 3 Remote Synchronization
- 4 Similar/Prior efforts
- 5 Use Cases
- 6 Design Options
- 7 Possible Back-ends
- 8 Indexing
- 9 Support for other Remote Tools besides Build
- 10 Milestones
- 11 Additional Features
This document attempts to describe the current state of the ideas and designs for a synchronization based file access for remote projects. This document has been started by Roland Schulz based on discussions on bug 316709.
Remote project support in Eclipse (i.e. projects where the source code is located remotely, and the executable is built and run remotely) is an essential requirement for science and engineering applications. Currently, options for using remote projects in Eclipse are:
- Network filesystems
- Revision control systems
- Remote services
The problems with these approaches are discussed below. The approach being proposed has different disadvantages (see below). Thus this approach is not meant to replace the current RDT approach but offer an alternative. This will allow the user to choose the approach with those advantages best aligned to his working environment. Also this approach will reuse parts from RDT (e.g. scanner/indexer) from RDT.
Using a network filesystem, such as NFS, to trick Eclipse into thinking the project is local suffers from a number of problems. The indexing and advanced editing features of CDT require accessing all source files in the project. These features already suffer performance related issues when accessing local files (particularly for large projects). When using a network filesystem, they can become unacceptably slow. In addition, network filesystems typically require system administration access to configure locally (although systems such as FUSE can be used to circumvent this) and on the remote system. This technique also requires a continuous network connection in order to access the project.
Revision control systems
Eclipse provides exemplary integration with a variety of RCS's, including CVS, SVN, and git. One approach to remote projects is to use the features of a RCS, to synchronize with a remote copy of the project. This has the nice feature of already being well integrated with Eclipse, but activities such as building and launching do not support this model since they need to take place on the remote system. In addition, there are other issue with scanner discovery and the remote environment that will be discussed in more detail below.
The Remote Development Tools (RDT) takes the approach of identifying the range of services required for C/C++ projects and creating remote implementations of these. These services can be broken down into the following categories:
- file access (for editing)
- managed build
- make build
- model builder (e.g. outline view)
- call hierarchy
- type hierarchy
- content assist
- include browser
- navigation (e.g. open declaration)
RDT currently supports either RemoteTools and RSE for providing remote services. Both Remote Tools and RSE use a remote DStore server to provide remote implementations of most of these services. File access is provided using the EFS abstraction in Eclipse. EFS services are provided either via the DStore server for RSE, or a separate SFTP service for Remote Tools (for historical reasons).
This approach has several disadvantages:
- Although caching can be used in many cases, responsiveness of UI is determined by network speed.
- A continuous network connection is required for project access
- A large number of users could overload the remote system with indexing
- Not all CDT functions can be supported using this model (e.g. refactoring)
- Remote search currently only works with RSE
- Not all Eclipse functionality fully supports EFS (e.g. Team services)
- Only CDT (i.e. C and C++) is supported, other languages, such as Fortran, would require a significant engineering effort
The Eclipse core and CDT are doing most of the file operations in the main thread (based on the assumtion that all file operations are low latency). This causes a responsiveness problems with a remote file system.
Because the file operations are in the main thread they block the GUI until the IO operation finishes and thus preventing the user to continue the work while the IO operation is running. It also often prevents IO operations which could run in parallel to do so. See Bugs 160353, 177994, 195997, 218387, 219169 and wiki.eclipse.org/TM_and_RSE_FAQ from the RSE team regarding the same problem for RSE. Their seems to be no work-around for this problem. While it seems in theory to be possible to improve it somewhat by using Display.readAndDispatch, it is not advised and has been removed from RSE (160353). Having a responsive UI is considered by many extremely important thus this is an important point.
It is very unlikely, at least for the medium-term (meaning the next Eclipse release in 2011), that both Eclipse Core and CDT move all file operations into threads and hide latency by doing IO operations in parallel. Therefore a different approach is needed to have an acceptably performing remote IO method.
As an alternative to the current techniques, we are proposing a new approach for remote projects. This approach relies on maintaining synchronization between two copies of the project: a local copy that exists in the user's workspace; and a remote copy that is used for building and launching the application.
- A local copy of the project exists in the filesystem, so network latencies are minimized
- Offline operation is possible
- All CDT functions are supported
- Other languages, such as Fortran require minimal effort to support
- Eclipse features, such as Team support, can also be used
- The entire project must be copied to the local machine. This only happens once, but could take a very long time for large projects/slow connections.
- Local indexing is problematic as the local environment will be different from the remote environment, so macros and includes will be incorrect. Running scanner discovery remotely seems to be the obvious way to solve the macro problem, but scanner discovery is hopelessly broken and not even the CDT people seem to know how it works. In addition, the indexer would need to be modified to copy system and library includes from the remote machine as part of the indexing.
- Some activities, such as building, will always need to be done remotely, so the performance problems will always be evident to some degree.
- rsync file subsystem Bug describing ideas to implement rsync based back-end for RSE
- Use rsync to sync the remote workspace to the local machine FAQ including disadvantages and advantages of using Rsync for a remote project
- PTP/photran/rsync remote projects wiki page describing the working Photran test implementation
- Rsync-style Remote Project Support for Photran
- Remote Include Path Support
The current implementation is a strawman prototype of a rsync-based remotely-synchronized project. It adds a new project wizard which creates a C/Fortran project but replaces the standard CDT build command (make) with a call to a custom shell script which uses rsync to copy the project to a remote server and run make remotely. This was definitely a prototype -- I'm sure the final version won't look anything like it (e.g., our build script makes two or three separate connections to the remote machine) -- but this is *simple* and it works, more or less, which gave us something real to try out.
Remote Include Path Support adds remote (Fortran) INCLUDE paths to Photran. Photran's include paths are configured in the project properties. Traditionally, they'd be paths on the local maching (e.g., /usr/include:/usr/local/include). This replaces them with URIs, so they can be on either the local machine or a remote one (e.g., rse://remotehost/usr/include:file:///usr/include). It also changes the properties page to use a remote file selection dialog box.
Outside from Eclipse
Synchronization using GIT
- website synchronization
- php web deployment using git
- deploy php using git
- 5 tips for deploying sites
- update website with a single command git push
Remote file systems
- User creates a new project (managed or make based) using "New remote sync project" wizard. User supplies the project name, remote host, username, password, and remote path. Empty project (or initial template) created on local and remote machines. Wizard sets up builder/scanner discovery/remote paths to work on remote system.
- Existing local project (managed or make based). Use "Convert to remote project" wizard to create a copy of the project on a remote machine and maintain synchronization between the local and remote copy. User supplies the remote host, username, password, and remote path. Wizard sets up builder/scanner discovery/remote paths to work on remote system.
- Existing remote project (make based). Use "Import remote project" wizard to create a local copy of the project and maintain synchronization between local and remote copy. User supplies the project name, remote host, username, password, and remote path. Wizard sets up builder/scanner discovery/remote paths to work on remote system.
- Existing local project that has been checked out of revision control system. Use "Convert to remote project" wizard to creates a copy of the project on a remote machine and maintain synchronization between the local and remote copy. User supplies the remote host, username, password, and remote path. Wizard sets up builder/scanner discovery/remote paths to work on remote system. RCS control files are not synchronized.
There are a number of design options that need to be addressed. These include:
- Should the synchronization be a project type? A project nature?
- When should synchronization occur?
- Synchronizing before any remote operation (build, remote index, ...)
- Synchronizing after each save
- Should synchronization be manual or automatic?
- How does synchronization play with Team support?
The 2nd option shouldn't wait on the sync but do it asynchronous. Otherwise the responsiveness problem (see above) wouldn't be addressed. Each remote operation would need to call a function to guarantee that all outstanding synchronization calls have finished. The same function would initiate the synchronization for option 1. Which option is better depends on the synchronization back-end and the user preferences and should thus be configurable.
Advantages of Sync after each save:
- Required for Auto-Build and Indexing on Server
- Reduces time to build (because is already synced)
- Causes larger repository (~2k per commit) and more traffic
Rsync and GIT both have advantages and disadvantages. Thus it sounds reasonable to support at least those two and make it easy to add others later on.
Advantages of Rsync
- widely available. And thus easy to set-up without installing software on the remote machine
Disadvantages of Rsync
- no JAVA implementation is available
- the synchronization is only one-way
- problems if clocks are not synchronized between systems
The 2nd is important if, either automatically or by the user, remote files get changed. The one-way synchronization of rsync would usually not synchronize changes to the client and would not detect conflicts caused by changes on both sides very well.
Advantages of GIT
- It has a java implementation (shipping with Helios)
- is known to be extremely fast (including the java implementation)
- supports two way synchronization.
Of course GIT is not meant as a synchronization tool (but a DCVS) but it works as a synchronization tool extremely well. Using git for synchronization would work both for those users using it also for version control and for those users using some other tool for version control. As an example a remote synchronization of a folder containing ~4000files (1 changed - which unknown to GIT), ~100MB, where GIT detects file changes on both sides, over a remote connection (cable), takes less than one second. The performance is mainly limited by the file system for the tree traversal.
Implementation Issues with GIT
To push to a non-bare repository is discouraged
Their are different options
- Fetch (not good option because it requires SSHD on the client)
- Push into working branch with post-update hook. Disadvantages: Requires
stat of each file on server (slow over NFS) and doesn't allow merge on client side
- Push to separate bare repository. Disadvantage: Requires 2 repositories
- Push to remote branch. Seems best option
Local indexing should be supported. Should we also support remote indexing? Local indexing requires remote include files. See above for the support for Phortran. For CDT it has to be checked how feasible it is to support remote include file paths.
Support for other Remote Tools besides Build
- Check feasibility of remote include path support in CDT (see PTP/designs/remote/sync#Indexing)
- Define a new Synchronization service type (which add synchronization/replication to the running EFS). It would have as public method guranteeSynchronized. The default server (for a purely local project or for remotetools/RSE) would do nothing.
- Add to all remote operations (compile, remote index, ..) a call to gureanteeSyncronized
- Implement a GIT based synchronization service (doing the GIT push in the gureanteeSyncronized call)
- Add the GUI to configure the synronization service (including new project wizard)
Later an EFS which would do the asynchronous GIT push after a file modifcation(e.g. save) and the gureanteeSyncronized would just wait for the push to finish.
Remote file view
To run the binary it is required to select it on the remote machine. For that it would be nice to be able to browse the remote machine using RemoteTools/RSE. Preferable the path shown is the path to the remote copy of the local project. Also it would be nice to have a "Project Explorer" View for the files on the remote machines. This would allow to view binaries, object files and other remote files not part of the synchronization. This view should also use RemoteTools/RSE
Build local and remote from same project
If the remote build is implemented as a builder which can be added to a standard CDT/Photran project, than it is possible to have both a local and remote builder for the same project. CDT already supports several builder configuration. Thus this can be used to switch between local and (potentially several) remote builder. It has to be checked that the indexer (including the remote include files) is updated correctly when the builder configuration switches.