Skip to content

Instantly share code, notes, and snippets.

@marcesher

marcesher/state_is_a_bug.txt

Created November 9, 2012 14:29
Download ZIP

Revisions

  1. marcesher revised this gist Nov 9, 2012. 1 changed file with 1 addition and 1 deletion.
    2 changes: 1 addition & 1 deletion state_is_a_bug.txt
    View file
    Original file line number Diff line number Diff line change
    @@ -1,6 +1,6 @@
    Joe, thanks for responding.

    I'm most interested in what are perhaps pedestrian issues, but they're issues nonetheless. You mentioned "outsourcing" session management... where can I read more on that?
    I'm most interested in what are perhaps pedestrian issues, but they're issues (for me) nonetheless. You mentioned "outsourcing" session management... where can I read more on that?

    Take the simple case of: I'm a user on your system. I'm logged in. I'm doing things. The server I'm on disappears while the screen I'm reading is currently loading.

  2. marcesher created this gist Nov 9, 2012.
    13 changes: 13 additions & 0 deletions state_is_a_bug.txt
    View file
    Original file line number Diff line number Diff line change
    @@ -0,0 +1,13 @@
    Joe, thanks for responding.

    I'm most interested in what are perhaps pedestrian issues, but they're issues nonetheless. You mentioned "outsourcing" session management... where can I read more on that?

    Take the simple case of: I'm a user on your system. I'm logged in. I'm doing things. The server I'm on disappears while the screen I'm reading is currently loading.

    What happens? Do I see an error? Am I sent to a new machine, with all my session state in tact, and the screen I was reading simply reloads?

    Or take perhaps a different architecture where a machine isn't brought down until all its users have been successfully moved to other servers. For example, perhaps our configuration is such that we have N instances, and we do a rolling code push to those instances. User A is on Server 5, which has the old code. We need to get that user to Server 2, which has the new code, along with all of his state. Once all users are off of Server 5, the deploy then moves to Server 5 as well and then Server 5 is brought back into commission and accepts new requests.

    So: what techniques, processes, and tools support these practices?

    Thanks again.