|
Jamming
Guest
|
 |
« on: January 17, 2011, 01:34:49 PM » |
|
I don't know for sure. The server is still there, it just seems that it's won't accept connection attempts, as it pings just fine and identifies itself. Either that or it reached its bandwidth limit, which I doubt. If either happened it would seem that this is one popular Episode.  I just hope it isn't anything to do with that orgy scene after the dinner.   |
|
|
|
|
Logged
|
|
|
|
|
nlowell
|
|
« Reply #1 on: January 17, 2011, 02:14:19 PM » |
|
I don't think it was us. They've been having hardware issues for a while, I think.
Just looks like it might be something deeper than a wonky drive.
|
|
|
|
|
Logged
|
|
|
|
|
Jamming
Guest
|
|
« Reply #2 on: January 17, 2011, 02:29:22 PM » |
|
Oh fine, just ruin it for me. Here I was thinking you were so popular that we would have to hire you a bodyguard. |
|
|
|
|
Logged
|
|
|
|
|
Laith
|
|
« Reply #3 on: January 17, 2011, 02:41:58 PM » |
|
I've been talking with Evo over at podiobooks. He says the issue is a combination of hardware and the webserver software.
And while Nate does have a nice big fan base the incremental load from us while not helping isn't causing the problems.
|
|
|
|
|
Logged
|
- Sitting down to attempt to listen to Ravenwood again.
|
|
|
|
mimccart
|
|
« Reply #4 on: January 17, 2011, 02:52:34 PM » |
|
They need to shutdown the service while they work, set up a temp server to provide an "Under Maintenance" page, or change the DNS entry and get a friend to host a message home page until they finish. I can't imagine that all the hits and communication attempts is good for the bandwidth or processor while they are trying to get things worked out. Most users probably don't know they are down hard and will just keep hitting the site...just like we did until we were told they had a bigger problem.
|
|
|
|
|
Logged
|
The Iris can sail circles around the Lois McKendrick and still be back at the Orbital for the afternoon deals at the flea market.
Quis custodiet ipsos custodes?
|
|
|
|
Laith
|
|
« Reply #5 on: January 17, 2011, 03:00:02 PM » |
|
They need to shutdown the service while they work, set up a temp server to provide an "Under Maintenance" page, or change the DNS entry and get a friend to host a message home page until they finish. I can't imagine that all the hits and communication attempts is good for the bandwidth or processor while they are trying to get things worked out. Most users probably don't know they are down hard and will just keep hitting the site...just like we did until we were told they had a bigger problem.
I think they have shut down apache now while they work on it. That is why the machine pings but no pages get served. getting a friend to host is more complicated than you would think, it requires messing with the DNS as you said and you DON'T want to do that if you can avoid it. |
|
|
|
|
Logged
|
- Sitting down to attempt to listen to Ravenwood again.
|
|
|
|
nlowell
|
|
« Reply #6 on: January 17, 2011, 03:22:43 PM » |
|
community.podiobooks.com is not on the same box and is a good "go to" site for info
|
|
|
|
|
Logged
|
|
|
|
|
Jamming
Guest
|
|
« Reply #7 on: January 17, 2011, 10:00:18 PM » |
|
See ya did break it, Mister "15% of total Bandwith" Lowell.  |
|
|
|
« Last Edit: January 17, 2011, 10:03:53 PM by Jamming »
|
Logged
|
|
|
|
|
Laith
|
|
« Reply #8 on: January 17, 2011, 10:14:15 PM » |
|
See ya did break it, Mister "15% of total Bandwith" Lowell. Evo must have taken a second look at things, he had told me that we weren't that significant an additional load. and since most of us don't download more than once I would go with this being a proud indicator of your success Nate. I think if the hardware hadn't had issues waiting to happen it probably would have been fine. |
|
|
|
|
Logged
|
- Sitting down to attempt to listen to Ravenwood again.
|
|
|
|
Tara_Li
|
|
« Reply #9 on: January 17, 2011, 11:53:45 PM » |
|
getting a friend to host is more complicated than you would think, it requires messing with the DNS as you said and you DON'T want to do that if you can avoid it.
The normal approach I've seen is to set up a webserver on a different machine, whether it's on the primary interface, or on a secondary interface, that responds to the same IP, and have it serve up a static page for all requests to minimize load, then drop the old box and let the ARP tables on the local wire update and move on. That way, no need to fudge DNS records. |
|
|
|
|
Logged
|
|
|
|
|
mimccart
|
|
« Reply #10 on: January 18, 2011, 12:19:49 AM » |
|
getting a friend to host is more complicated than you would think, it requires messing with the DNS as you said and you DON'T want to do that if you can avoid it.
The normal approach I've seen is to set up a webserver on a different machine, whether it's on the primary interface, or on a secondary interface, that responds to the same IP, and have it serve up a static page for all requests to minimize load, then drop the old box and let the ARP tables on the local wire update and move on. That way, no need to fudge DNS records. Okay, but it requires that you have a backup piece of hardware available. How do you get another interface to respond to the same IP address? Doesn't that violate the basic rules of TCP/IP? Or are you describing something more like content level load sharing? Normally I would have guessed that a site this extensive would have several identical servers load sharing or alternatively set up as the html server, app server, storage server, etc. I also would expect a couple of raid 5 or better jbods for storage and redundancy. Maybe they did, but the commentary makes it sound like there really was a single point of vulnerability, and thus, failure. With that in mind, setting up a temp front page on someone else's server and making that one line NAME change in the DNS entry is trivial compared setting a separate box, interface, web server, etc. It also has the added value of offloading traffic almost immediately after making the change. All of it seems simpler to me based on my interpretation that they have a simple set up. DNS would be nearly a no-brainer under those circumstances. If the set up is actually more complicated, then things could ugly on a lot of levels. |
|
|
|
|
Logged
|
The Iris can sail circles around the Lois McKendrick and still be back at the Orbital for the afternoon deals at the flea market.
Quis custodiet ipsos custodes?
|
|
|
|
Jamming
Guest
|
|
« Reply #11 on: January 18, 2011, 07:22:17 AM » |
|
Nate, What is the plan for updates until Friday? You have any idea what you want to do after Episode 22?  |
|
|
|
|
Logged
|
|
|
|
|
nlowell
|
|
« Reply #12 on: January 18, 2011, 07:24:37 AM » |
|
As soon as I get out from under damage control here, I'm gonna edit 23 and post it on my site.
That should happen later today.
|
|
|
|
|
Logged
|
|
|
|
|
Jamming
Guest
|
|
« Reply #13 on: January 18, 2011, 07:31:06 AM » |
|
As soon as I get out from under damage control here, I'm gonna edit 23 and post it on my site.
That should happen later today.
Thank you for the prompt response, now I can sleep with a relaxed mind and rest my weary brow. |
|
|
|
|
Logged
|
|
|
|
|
Tara_Li
|
|
« Reply #14 on: January 18, 2011, 07:55:22 AM » |
|
Okay, but it requires that you have a backup piece of hardware available. How do you get another interface to respond to the same IP address? Doesn't that violate the basic rules of TCP/IP? Or are you describing something more like content level load sharing? Not at all. For example - Machine A is media.podiobooks.com, x.x.x.10 and Machine B is normally community.podiobooks.com, x.x.x.11. In both cases, those IPs are configured on interface eth0. Now, you can either 1) have a second network card in Machine B (a very common case for servers, and not uncommon for home computers now) that shows as eth1, or 2) set up a virtual interface that shows up as dummy0, and either of those can have x.x.x.10 attached to them. You get it configured, but not brought up - then you shut down Machine A, which will then no longer be responding to the router asking which machine is handling x.x.x.10, and bring up the interface on Machine B, which *will* respond to that request, so the router will then send that traffic as well to the new location. This is done very transparently, with the routers handling the issues. This makes use of the ARP (Address Resolution Protocol), which is hidden behind a router and never leaves the local network (local network being the collection of machines behind the router running the BGP protocol, which will take longer to update... but not as long as DNS by any means). |
|
|
|
|
Logged
|
|
|
|
|
