I'm rearranging the server layout.
Anyone paying close attention will notice some time over the next day, that if they reconnect, they will move from pointing to LasVegas, to pointing to LosAngeles.
It's a new front end server for the pool and will simplify the next change.
That being: the back end pool will be moving in the next 2 days.
When I move the back end there is likely to be a mining outage of up to 5 minutes.
I'll post more about the changes before they happen.
If you mine to nodes, you'll see no changes.
If you mine to the main pool in LasVegas, you'll see a change to LosAngeles.
The difference is about 8ms between the 2, so nothing major (or slightly better depending on where you are)
Edit: and in case anyone was wondering - it's an upgrade, not a down grade, a better back end server and 2 more servers in front of it
Kano, how long are you leaving the old server in Vegas up for? I have a bunch of things that are hard coded to point to only IPX and need to change everything to IPY before that old Vegas server is taken down. I like to lock everything down as much as possible.
Using the IP address is a bad idea for anyone.
The new IP address doesn't point to the new server, just to a front end passthru (which uses truck loads of CPU even though it does nothing else, and reports zeros for everything, I guess no one actually ever tested/used it including the developer
)
Thus if ever something drastic happens (e.g. like an extended DDoS, which hasn't happened before, or an extended server/provider failure) it's a minor change to move the front end and change the IP address via DNS.
Soon after I complete the changes I'll contact a few of the large miners and give them a 2nd front end whitelisted to allow a second (similar) access.
(i.e. I'll only allow access to that server from a set of IP addresses)
However, of course, there's no guarantee that one could never be DDoSed either, since a large DDoS can affect a whole data center, and of course server failures are always possible.
A redirect to point miners to another frontend, if an outage is known in advance, will fail if you mine to the IP address.
Redirect in cgminer works based on matching your current connection with the redirect, if they don't match, the redirect will be ignored.
e.g. stratum.kano.is => zomg.kano.is would work, but IPADDRESS => zomg.kano.is wont.
You could use a passthru at your data center, and point all your miners to that and then only need to change that if ever anything needed changing.
(Edit: or a straight proxy since that uses a lot less bandwidth and thus means slightly faster block changes for your miners when you have a lot of miners - usually setup as multiple proxies if you have lots of miners, so your miners can failover among them if one ever fails/restarts/whatever, but also, of course, have a final pool failover in the miner pool list)