Comments on: "Let It Crash" Programming http://blogs.teamb.com/craigstuntz/2008/05/19/37819/ C# • Entity Framework • Functional Programming • MVC • Web Sun, 12 Feb 2012 04:49:16 +0000 http://wordpress.org/?v=2.6.2 By: Seven Languages in Seven Weeks | GET http://localhost…200 OK http://blogs.teamb.com/craigstuntz/2008/05/19/37819/#comment-30062 Seven Languages in Seven Weeks | GET http://localhost…200 OK Tue, 07 Jun 2011 20:07:52 +0000 http://blogs.teamb.com/craigstuntz/2008/05/19/37819#comment-30062 [...] Erlang, but just reading the chapter and following the examples made it clear how the “let it crash” philosophy together with language support simplifies the process of writing robust [...] [...] Erlang, but just reading the chapter and following the examples made it clear how the “let it crash” philosophy together with language support simplifies the process of writing robust [...]

]]> By: miguel rodriguez http://blogs.teamb.com/craigstuntz/2008/05/19/37819/#comment-3566 miguel rodriguez Tue, 20 May 2008 17:45:00 +0000 http://blogs.teamb.com/craigstuntz/2008/05/19/37819#comment-3566 for years, I tried to make systems without failures, and I failure, the very day I accepted the failure and I worked on how to handled it, it suddenly all became stable and robust, because at least from the client perspective it doesn't matter what is behind the wall, I know this mental model it is not appreciate as It should be, but everything will be appreciate in time, I have no doubt about it. for years, I tried to make systems without failures, and I failure, the very day I accepted the failure and I worked on how to handled it, it suddenly all became stable and robust, because at least from the client perspective it doesn’t matter what is behind the wall, I know this mental model it is not appreciate as It should be, but everything will be appreciate in time, I have no doubt about it.

]]> By: Stephen Hoffman http://blogs.teamb.com/craigstuntz/2008/05/19/37819/#comment-3565 Stephen Hoffman Tue, 20 May 2008 16:33:24 +0000 http://blogs.teamb.com/craigstuntz/2008/05/19/37819#comment-3565 The OpenVMS Distributed Lock Manager has offered analogous capabilities for many years, though the application itself is expected to fully participate in the cooperative processing. If the application or the host within the cluster crashes, then DLM frees up the (arbitrary) locks (formerly) held by the failed application (or failed host), and cooperating applications are then granted the locks and can detect the failure and can take failure-appropriate action. This application processing can range from a full restart to specific recovery processing, to transactional-style sequencing. It's all fully programmable. Further, cluster locks can easily be used akin to the C wire that was described; where acquiring a lock can cause all processes monitoring the lock to take an application-specific action. A block of arbitrary and application-specific information can also be passed along with the lock itself. This could be a next invoice number, or a count of crashes, or information related to the failure or the recovery and restart; this is an entirely arbitrary block of data. The OpenVMS Distributed Lock Manager has offered analogous capabilities for many years, though the application itself is expected to fully participate in the cooperative processing.

If the application or the host within the cluster crashes, then DLM frees up the (arbitrary) locks (formerly) held by the failed application (or failed host), and cooperating applications are then granted the locks and can detect the failure and can take failure-appropriate action.

This application processing can range from a full restart to specific recovery processing, to transactional-style sequencing. It’s all fully programmable.

Further, cluster locks can easily be used akin to the C wire that was described; where acquiring a lock can cause all processes monitoring the lock to take an application-specific action.

A block of arbitrary and application-specific information can also be passed along with the lock itself. This could be a next invoice number, or a count of crashes, or information related to the failure or the recovery and restart; this is an entirely arbitrary block of data.

]]>