> > at this time. > [/QUOTE] > > However, my app would still not start at Mar 14, 2010 - 4:25 UTC. >
That sounds odd for sure. In the future, make sure to submit a ticket if you're seeing errors. We had no other reports that I'm aware of. It's possible you got unlucky somehow. > So, what exactly is the definition of "New and idled apps". My > understanding is that as long as I'm paying for dynos that my app will > not go idle, but if I increase the dynos from say 4 to 10 are the 6 > new dynos considered 'new' and won't start in such a situation? What > about workers? > What about accidentally pushing a 'new' version of code during such an > outage? I assume this would effectively bring my running app offline. > When issues like this occur, the system automatically identifies it, and prevents users from hurting things. You can't push, you can't change your dynos or workers. We make sure your app continues to run while we find and resolve the issue. > > 4) If I have a running app that is being concurrently hit by say 100 > users. What is the exact process the server takes when I push a new > version of code? In other words, do those 100 users immediately get > disconnected, or does Heroku spin up the new app, redirecting all new > requests to the new version once it is online, and then terminate the > 'old' version dynos after all existing requests to those dynos have > been satisfied? Basically, what is the user experience for someone > who is in the middle of submitting their credit card info at the > precise time that I push a new version of code? > The latter. Open request are served on the old dynos, which are then replaced with new ones. Users don't see anything. > > 5) I'm a little concerned about recent posts dealing with getting > large datasets in and out of the database on Heroku. I've read plenty > of creative ideas on how to get the data in and out, but nothing that > would really be considered acceptable for a high traffic site that > needs virtually zero downtime (especially with the posts saying that > Taps gets really slow after about 500 MB). Does Heroku have any plans > of making it easier to get large datasets in and out of the database? > What about scaling the database? It's easy to spin up app dynos, but > what about moving from a Fugu database to a Zilla? What kind of > downtime are we looking at for such a transition? (How about from > Ronin -> Fugu?) The faqs state: "Switching to/from a dedicated > database usually takes one business day as our support staff processes > the requests, and verifies that data migrations completed > successfully." I assume this is for switching between say Koi and > Ronin, but is it also true for Ronin->Fugu->Zilla? > There's two issues: one the ingress/outgress of big data. We're working on improvements and docs for this. Two - dedicated DB migrations. We handle this for you, and will coordinate with you. It takes 1-2 days to do, with a few minutes of down time each time you switch. That's between any dedicated option. We'll work with you directly to coordinate. > 6) I assume I would be looking to move away from Heroku long before I > reached 2TB of data limit on the database (since at that point I would > be looking for redundant systems that can have read/write traffic > separated, etc which Heroku doesn't appear to support), but how would > I get this data off Heroku? > > We haven't had anyone leave due to traffic or size yet. If you start hitting performance bottlenecks, we'd love to work with you to optimize the system or make any changes needed. That said, you can always request a data dump for us if the tools aren't working for you. 7) What systems are used to store the database data? In other words, > how safe is my data in the event of system failure (hard disk, etc)? > Single disk, Raid 1, Raid 5, Raid 10, etc? > Very safe. There's a combination of protection at every layer. SW raid across EBS (which are themselves RAIDed), daily backups of all data, etc. > 8) In the event of system failure, what kind of support can I expect > from Heroku to get the database back online? (Meaning that a hard > disk failure with a response of "restore from your backup bundle" > would not be considered acceptable. Transaction logs would need to be > replayed, etc.) > > We provide full support for this. We'll take care of it for you. > 9) DNS related: Is it possible to have say 2 set subdomains for a > given domain (ex: forums.mydomain.com, chat.mydomain.com) that point > to separate apps, and then have a catchall domain point to a 3rd app? > Ex: subdomainXYZ.mydomain.com points to a given app for any value of > XYZ. The docs seem to indicate that I need to use the heroku console > to add any subdomains, so what I'm looking for is how I would add a > catchall domain in Heroku. > > Wildcard domain, or add each domain independently, your choice. Oren -- You received this message because you are subscribed to the Google Groups "Heroku" group. To post to this group, send email to [email protected]. To unsubscribe from this group, send email to [email protected]. For more options, visit this group at http://groups.google.com/group/heroku?hl=en.
