Planning for unplanned downtime

Image representing Twitter as depicted in Crun...
Image via CrunchBase

Via D7: The Twitter guys speak | Beyond Binary – CNET News:

7:37 p.m. PT: They open it up for questions. Venture capitalist Roger McNamee offers a couple of comments. “Don’t ever do another planned maintenance in the middle of the day on a week day.”

I remember the day Roger McNamee is talking about, when Twitter posted on the homepage that there would be some ‘planned downtime’ in the middle of the day.

In datacenter operations language, ‘Planned downtime’ in the middle of the day is really  unplanned downtime.

I’m not privy to what went on that day at Twitter, but my guess as an ex-datacenter guy is that there was some production issue affecting some number of users which a) would have got worse over time or b) was not yet an issue, but would have become an issue had they not taken down the site at that time.

From a PR perspective and from a tech perspective (assuming my assumption from the previous paragraph was correct), Twitter did the right thing.  I can certainly appreciate a high growth site like Twitter having some growing pains (having been through that myself at multiple high traffic internet properties) and this was probably the best way to handle it.

Reblog this post [with Zemanta]

Edit

Leave a Reply

You May Also Like

Back on the couch

Scott just posted an amazing interview with Jerry Colonna. Felt like I was back on the couch. Best part of…

Just Make It Faster

The final myth is that cloud computing will make all performance and scaling issues go away. I heard…