|
[Date Prev]
[Date Next]
[Thread Prev]
[Thread Next] Re: scheduled power outage for NPB (fwd)
This is shinoda ... Okay, here's my action list. 1. I will Install a news system (preferably inn) on japan.park.org and our staging server. Verify that my scheme works between these two (almost 100% positive about this.) If successful, I'll feedback my configuration for step 3 below. 2. Chris should run a final test with his new guestbook system, and throw it in the town as soon as possible to verify it's correct operation under the real environment. 3. In the mean time, the news system should be installed in all park systems. (I'm not sure about this... should we just look at May 25 for now and limit changes to park.org (town) and japan.park.org?) 4. Modify the new guestbook system like as follows: (a) Split a 'POST' processing part into two part. (b) A new comment is 'posted' to local nntp server at some point, probably before either an actual storage or an indexing takes place. This is as far as the first part of the split up part works. (c) Assuming the comment is delivered through NetNews, make the latter part of the 'POST' processing as a separate program, taking its input from a file (a path into a news spool will be given) or standard input, do the rest of the work. 5. It might be a good idea to run a test 'in parallel' to the original new system (make a simple script that calls the original new system and modified new system simultaneously). The NetNews (inn) served us without a glitch throughout the 42 hours of operation handling more than 6,000 personal safety records in the Internet disaster drill as our database synchronization mechanism, including the first hour which we have intensionally dropped network connections to one of our four service sites. My brief is that using the news system as means of making (somewhat slow, but) coherent databases is a very robust approach. As I said, if cross feed is redundant enough, a chance of dropping a record entirely in transit is very small. In addition, we have our news spools as backup. We can periodically run a coherency test if we want (easiest way is to re-run the content of news spool again at all site). Database in physically crashed site may be restored in the same way. --- shinoda P.S. Is Chris on not-park? P.P.S. I can't examine Chris's new code. They are not mirrored nor directly visible under /Tools/Guestbook...
Follow-Ups:
[Date Prev] [Date Next] [Thread Prev] [Thread Next] |