newark issue...
20 Replies
@ianneub:
Don't worry, its a network problem, I'm sure your data will be safe.
I'm losing money :°(
It's pretty unusual for a data center to have connectivity problems that take this long to sort out.
@jcr:
I think it is time for linode to offer a dns failover between its datacenters.
Yes please. I'm very keen to dual host in London/Newark, and hoping the backup system will make it easy for me to maintain synch.
"Keen", in this case, should be read as "ready to throw money at Linode".
* If I had more than one node and in different datacenters, but only one node active at a time (web server, for example), if the active node's datacenter went down, couldn't I bring up the alternate node and change the DNS records in the Linode DNS Manager to point to my alternate node? And then be back in business once the DNS changes propagate? Of course, this assumes that the alternate node keeps in sync with the active node.
@TheHeartSmasher:
The issue appears to be resolved now.
So solved in about 2 hours…that seems reasonable.
Does Linode publish more information about the problem?
@mattm:
Yes this brings up a major concern over DNS failover. I would pay for additional nodes if I could get the DNS failover to operate between data centers.
It's there, and has been for some time.
Set the TTL nice and low, and failover should take <20 minutes from when your systems detect a problem; if failover needs to happen more quickly, you'll probably need to set up your own master nameservers and push updates that way.
@hoopycat:
It's there, and has been for some time.
~~[http://www.linode.com/api/index.cfm?method=domain.resource.update " target="_blank"> ](http://www.linode.com/api/index.cfm?met … rce.update">http://www.linode.com/api/index.cfm?method=domain.resource.update ](
Sure, in the same ways I could have run my own crons to do local backups. But I'd much rather pay to have it done for me. No admin, and somebody else to blame if it fails. What's not to like?
I don't even want to know my second box exists, never mind keeping it in synch. I just want it to automagically (and preferably immediately) pick up the load the next time Newark has network problems.
There are, of course, a myriad of contract sysadmins and developers who could make this work for your particular situation, but they aren't under the Extras tab
"Zones that require generation, activation/deactivation or deletion are done so every 15 minutes"
it would be nice to drop that time to a few seconds (ok, maybe 60 seconds).
@jcr:
From the Linode DNS Manager,
"Zones that require generation, activation/deactivation or deletion are done so every 15 minutes"
it would be nice to drop that time to a few seconds (ok, maybe 60 seconds).
I think the 15 minutes applies to the generation or activation/deactivation of the entire zone – ie., adding a new domain to the servers.
I run a DNS server on a Linode for work. My server is the master, and Linode's servers are slaves. When I make changes to my zones, they get pushed out to the slaves really quickly -- usually within a few seconds.
I know it's a small niche service, but I'm a really big fan of Linode's DNS. I can edit my zone files myself on my vps, with all of the flexibility that brings, and I get the power and reliability of a good network of servers.
I believe that if you have your own DNS server, and have Linode's ones as slaves, they support push-notification and update immediately after receiving such.
@rsk:
Changes done via DNS Manager, even single-entry zonefile edits, are pushed to the servers four times per hour. So, if you edit at xx:10, you need to wait only about five minutes.
I believe that if you have your own DNS server, and have Linode's ones as slaves, they support push-notification and update immediately after receiving such.
You are correct.
@phine:
Another incident (network issue) in Dallas data center. My site was down for an hour. And whenever Dallas center has a hiccup, linode.com is down (really bad). Hopefully this does not become a trend.
They publish information here