Stuff to get it right early for a startup

-
I'll make this a short post. It takes less time to set this up initially and get all of your projects conform to it than trying to retro-fit it later. And the effort spent on this will pay for itself in saved time from increased productivity ten times over.

  • A source control repository
    • Separate binary file assets (like lots of images, videos etc) from text file assets (like source code into separate repositories.
    • Use a distributed version control repository, like git.
  • Integrated Code review tool, like gerrit.
  • Integrated Bug database, like bugzilla (it's very customizable and fast) or jira (newer versions are pretty good).
  • Integrated code browser, like opengrok. 
  • Every project should be buildable, preferably using autotools.
    • Even if it's 3rdparty code, never just keep the binary. Always keep the source in good building shape.
    • Also, save the web url or location from where it was downloaded. There maybe a bugfix or a update you may want to pick up later.
    • Ensure build is fast. Use distcc and ccache to make it faster.
    • Split the overall code into independent layered one-way dependency projects.
  • Continuous build and deploy and smoke-test setup
    • This is extremely important for a project that's in active development.
    • Ensure smoke test is up-to-date and extensive.
    • Build system itself should be version-controlled. Treat build systems as sacrosanct. Don't install or upgrade packages randomly on this system.
  • Don't be stingy about the hardware for these systems.
  • Backup everything.
For a startup, hiring a build-automation engineer who can do all this stuff correctly and efficiently is probably money well-spent because it will save tons of effort on behalf of your costlier engineers,  architects etc.

If you are building an Internet application, then there are more things to pay attention to - like ensuring your build system can publish to software distribution system, upgradable builds with build ids and version numbers and automatic dependency management etc. Maybe another quick post on it sometime later.

Popular posts from this blog

Linux IPC and Limits

-
Note: POSIX implementation provides better and cleaner api to IPC compared to System V. Prefer using POSIX api if you are known to be running on 2.6 or later kernel. System V IPC implementation on Linux includes Shared memory, Semaphores and Message queues. There are system imposed limits that are important to keep in mind when developing application/services in Linux. To know the limits use ipcs command. For example on my system: [root@f7 ~]# ipcs -l ------ Shared Memory Limits -------- max number of segments = 4096 max seg size (kbytes) = 32768 max total shared memory (kbytes) = 8388608 min seg size (bytes) = 1 ------ Semaphore Limits -------- max number of arrays = 128 max semaphores per array = 250 max semaphores system wide = 32000 max ops per semop call = 32 semaphore max value = 32767 ------ Messages: Limits -------- max queues system wide = 16 max size of message (bytes) = 8192 default max size of queue (bytes) = 16384 Or via sysctl interface: kernel.shmmni = 4096 kernel.shmal
Read more

htop: more than top. pstree: ps with tree.

-
Image
I'm a big fan of procexp and other sysinternals tools on Windows. Many people who use those and then come to Linux complain that there are no such tools in Linux. Most people are familiar with top and ps commands. I would like to introduce you to a bunch of utilities on Linux that are very useful while debugging issues. Lets start with htop, a better alternative to top. You can install htop on a yum based system like Fedora, Redhat, CentOS etc with "yum -y install htop" or with "apt-get install htop" on systems that are debian apt based like Ubuntu. htop is much better in displaying the process and resource information than top. It gives a more accurate and clearer picture of process memory usage than top. A screenshot of htop in action: It has many interesting features. It's display layout is refreshing different and nice. It can show tree view, a very important feature to understand the ancestry of a process. In the size column, the MB part is hig
Read more

Access Apache WebDAV from Windows XP

-
This is a work-in-progress. This is not complete, may be never will be. It is purely for my own reference. Use at your own risk. Here I try to remember how to configure Apache and Windows XP so that WebDAV on Apache is accessible from Windows XP clients. Links: http://www.thrrrust.com/~pclark/blog/2005/03/fun-with-windows-xp-and-webdav.html http://www.hss.caltech.edu/help/unix/admin/web/apache-webdav-xp http://support.microsoft.com/?kbid=909423 http://www.greenbytes.de/tech/webdav/webdav-redirector-list.html Registry changes required to support large files and many files: http://support.microsoft.com/default.aspx/kb/900900/ http://support.microsoft.com/?kbid=912152 Automatic Protocol Driver Selection for MSDAIPP http://msdn2.microsoft.com/en-us/library/ms691698.aspx http://msdn2.microsoft.com/en-us/library/ms691696.aspx http://msdn2.microsoft.com/en-us/library/ms691722.aspx DHTML Behaviors for making webfolder open automatically. http://dean.edwards.name/moz-behaviors/ http://msdn.micr
Read more