We have Sonic fiber in San Francisco, and about 2 months ago our household starting noticing that web pages and apps were loading very slowly. This typically happened the first time visiting a given site, and then future page loads were quick again. The slow loading occurs on web pages across two macbooks and a linux laptop, a couple Android phones, and a PC connected directly to ethernet, so it is not device specific and not just a wifi issue.
I am running the SR516ac router that Sonic provided. It's running firmware 2.6.2.8 (build 220610_1947), which to my understanding is also installed by Sonic. I have not messed with the config. Rebooting the router only fixes it briefly.
Signs point to a DNS issue, so I've been testing the DNS response time from the router. For all experiments I have a list of 30k domains that I ensured were good by checking that they gave a valid DNS result. I then sample a random domain and see the latency to get a DNS response back, or if the DNS server times out. I'm sampling a random domain from a large list so that hopefully the domain has fallen out of the local DNS cache.
Here's the baseline. I queried a random domain every 10 seconds for 8 hours. The blue bars show the time to a response, and the red X's show DNS failures:
5.8% of queries were failures, which seems high to me.
It gets worse. I found that certain queries were guaranteed to lock up the DNS server. For example, this domain no longer exists, but when you query it, it always takes the full amount of time to timeout, and the router's DNS server becomes unresponsive for ~15 seconds while it's being queried:
Here is what the DNS latency looks like while querying the poisonous domain name:
The poisonous domain was being queried in the background from about 18 to 35 minutes in, and you can see the corresponding DNS failures on the plot. Then the DNS server recovers for a bit, then fails a bit more, then recovers again.
Is there something further I can do to track this down or fix this issue? I'm not very familiar with networking, so I'm not sure if there's anything within my power to fix this, or if it's entirely a Sonic problem.
Thanks!!
I am running the SR516ac router that Sonic provided. It's running firmware 2.6.2.8 (build 220610_1947), which to my understanding is also installed by Sonic. I have not messed with the config. Rebooting the router only fixes it briefly.
Signs point to a DNS issue, so I've been testing the DNS response time from the router. For all experiments I have a list of 30k domains that I ensured were good by checking that they gave a valid DNS result. I then sample a random domain and see the latency to get a DNS response back, or if the DNS server times out. I'm sampling a random domain from a large list so that hopefully the domain has fallen out of the local DNS cache.
Here's the baseline. I queried a random domain every 10 seconds for 8 hours. The blue bars show the time to a response, and the red X's show DNS failures:
- plot_baseline_8hour.png (24.35 KiB) Viewed 910 times
5.8% of queries were failures, which seems high to me.
It gets worse. I found that certain queries were guaranteed to lock up the DNS server. For example, this domain no longer exists, but when you query it, it always takes the full amount of time to timeout, and the router's DNS server becomes unresponsive for ~15 seconds while it's being queried:
Code: Select all
dailyimmumax.comHere is what the DNS latency looks like while querying the poisonous domain name:
- plot_poisonous.png (22.97 KiB) Viewed 910 times
The poisonous domain was being queried in the background from about 18 to 35 minutes in, and you can see the corresponding DNS failures on the plot. Then the DNS server recovers for a bit, then fails a bit more, then recovers again.
Is there something further I can do to track this down or fix this issue? I'm not very familiar with networking, so I'm not sure if there's anything within my power to fix this, or if it's entirely a Sonic problem.
Thanks!!
