This past Monday evening we implemented a new load balancing mechanism for node registration known as RRDNS (round-robin DNS). This occurred without notifying our users and we apologize for doing it without notice.
While RRDNS accomplishes our major goal of distributing the registration load between our three servers, there may be some undesired side effects. Along side of the RRDNS development effort another option for load balancing was being developed. Like RRDNS, it uses the DNS layer to affect load-balancing, but it leverages Amazon’s AWS DNS service, Route53. Instead of round-robin distribution, it uses Route 53’s latency-based DNS distribution mechanism and is more compatible with AllStar nodes. Route 53 will register your node to one of our three servers and will not change unless there is a failure.
We are looking for experienced AllStar users to help us decide which method is best. If you’d like to help us test Route 53, the attached document provides detailed instructions how to modify your node. Please read them carefully and be sure your are comfortable making the changes.
Of course, we’d love to hear your feedback, good, bad or otherwise.
I don’t think us.reg.allstarlink.org is properly making latency decisions?? I’ve tried DNS tests from global locations and everything points back to 44.98.254.145. As an example, here are the results from Chicago…Pings show that 44.72.21.13 is closest:
mcgough@hvchicago2:~$ ping -c 5 44.72.21.13
PING 44.72.21.13 (44.72.21.13) 56(84) bytes of data.
64 bytes from 44.72.21.13: icmp_seq=1 ttl=56 time=6.79 ms
64 bytes from 44.72.21.13: icmp_seq=2 ttl=56 time=6.98 ms
64 bytes from 44.72.21.13: icmp_seq=3 ttl=56 time=7.06 ms
64 bytes from 44.72.21.13: icmp_seq=4 ttl=56 time=6.87 ms
64 bytes from 44.72.21.13: icmp_seq=5 ttl=56 time=6.82 ms
— 44.72.21.13 ping statistics —
5 packets transmitted, 5 received, 0% packet loss, time 4005ms
rtt min/avg/max/mdev = 6.795/6.907/7.066/0.114 ms
mcgough@hvchicago2:~$ ping -c 5 44.103.0.49
PING 44.103.0.49 (44.103.0.49) 56(84) bytes of data.
64 bytes from 44.103.0.49: icmp_seq=1 ttl=53 time=10.8 ms
64 bytes from 44.103.0.49: icmp_seq=2 ttl=53 time=10.6 ms
64 bytes from 44.103.0.49: icmp_seq=3 ttl=53 time=10.4 ms
64 bytes from 44.103.0.49: icmp_seq=4 ttl=53 time=10.5 ms
64 bytes from 44.103.0.49: icmp_seq=5 ttl=53 time=10.9 ms
— 44.103.0.49 ping statistics —
5 packets transmitted, 5 received, 0% packet loss, time 4006ms
rtt min/avg/max/mdev = 10.477/10.701/10.928/0.164 ms
mcgough@hvchicago2:~$ ping -c 5 44.98.254.145
PING 44.98.254.145 (44.98.254.145) 56(84) bytes of data.
64 bytes from 44.98.254.145: icmp_seq=1 ttl=54 time=35.0 ms
64 bytes from 44.98.254.145: icmp_seq=2 ttl=54 time=35.1 ms
64 bytes from 44.98.254.145: icmp_seq=3 ttl=54 time=34.6 ms
64 bytes from 44.98.254.145: icmp_seq=4 ttl=54 time=34.4 ms
64 bytes from 44.98.254.145: icmp_seq=5 ttl=54 time=34.4 ms
— 44.98.254.145 ping statistics —
5 packets transmitted, 5 received, 0% packet loss, time 4001ms
rtt min/avg/max/mdev = 34.409/34.752/35.190/0.329 ms
However dig ALWAYS returns 44.98.254.145 for us.reg.allstarlink.org. I’ve tried this from 8 sites around the USA.
Latency is used to determine which region ASW thinks you are in. That’s not necessary your least ping time to the registration server. There are two US zones; us-east goes to .145 and us-west goes .13. The rest of the world goes to .49. Of course, if there is a failure that would change.
That’s different than what I get. Here’s my server in LA.
Okay…I guess AWS Route 53 is using a pre-built geo-location database. Interesting that this site shows the same US results as I was getting (only resolving to 44.98.254.145):
I get the same results as you from that DNS Checker, all US sites resolve to .145. However there is only the one site in the west region.
I tested from a few of my other VMs and my Mac at home, all resolve to .13. I normally use Google’s DNS but I also tried 1.1.1.1 from a few places and those resolve to .13 as well.
Sounds good…My comments reflect more of a curiosity about AWS Route 53 than concern about operation. I’ll try nslookups from some IP addresses that geo-locate to other regions around the USA, out of curiousity.
Microsoft Windows [Version 10.0.17134.950]
(c) 2018 Microsoft Corporation. All rights reserved.
H:\>nslookup us.reg.allstarlink.org
Non-authoritative answer:
Name: us.reg.allstarlink.org
Address: 44.72.21.13
Sounds good…My comments reflect more of a curiosity about AWS Route 53 than concern about operation. I’ll try nslookups from some IP addresses that geo-locate to other regions around the USA, out of curiousity.
@kb8zgl That’s the correct IP for where you are. The issue @David_McGough is having is that everything in the US is resolving to .145 (TPA). Other than the DNSchecker, I resolve to .13 (ORD) all my US-west sites.
Well did notice something was going on, so did some looking around and finally got onto the site here. I used the check_reg.sh to see what servers my nodes were connected to. Is there any means to see what nodes are connected to a given server ??
Does check_reg.sh show you the IP address? If so there is a list of the hostnames in the attachment on the original post you can use to see which location you’re hitting.
There is an admin only page that shows us the registration server. I grabbed a couple screen shots and hid your IPs. Looks likes you’re using RRDNS because the registration server changes over time and in the second screen shot the registration server is different for nodes on the same server. If you switch to us.reg.allstarlink.org those side effects won’t happen.
