"Connection error (1.1.1.1#53): failed to send TCP(read_write) packet (Connection prematurely closed by remote server)"

General Beta 6.0
1

Hi,

the past days/week(s) im getting this "error". DNS Resolution still works, but i see this Warning/Error on my Dashboard always and it annoys me:

image
image1032×228 16.7 KB

https://tricorder.pi-hole.net/jsnfxVTo/

Thanks
Furkan

2

Despite the annoyance, this warning shows that communication with your specified upstream server wasn't possible. If this happens once, it may be irrelevant, if you have multiple TCP-capable upstream DNS servers, Pi-hole will try the next one. If not, the name resolution will fails and your client will (hopefully) retry the query. If, however, this happens more often, you may want to know about this and be able to intervene, e.g., by changing over to a "better" upstream server.

What would be your suggestion for improvement here?

3

Hi,
this happend a lot, very often even.
If i delete the warnings, 10sec later or 1-3mins later, i get the warning back again.
Tho i dont know why this happens randomly, since it never happend before.

but from what you telling me it just sounds like my Raspi has issues connecting to the upstream and it just retries the other upstream i have configured.

Since my internet works and i never see a downtime, i was just wondering why i am running into this.

4

Not necessarily. The warning your are quoting means connection has been established but the data transmission was not finished but instead, the connection was prematurely closed by the remote server. Most often, this comes from (very!) brief Internet connections outages. Out of interest, I set up a very tight connection monitoring between myself (FTTH) and three high-availability servers on the web and have been able to find hundreds of short outages throughout a day. They only last for less than a second each so us, human beings, we probably simply don't realize this at all in our daily lives. Maybe you see this once a week by chance that a webpage takes 0.5 seconds longer to load than usual and don't really realize what this means. The page eventually load as your browser will retry.

5

Aight thanks, i will just live with it or try different upstreams
Thanks!

6

@DL6ER
Its funny.. It happens to any other upstream i try. just randomly disconnecting, and the moment it fails to connect, it also just cant resolve any dns. I never had this issue

I kinda wanna exclude that i have internet interruptions because i can still ask my Pi without any issues.
Can i do other things to debug this?

7

If it happens to any server, the "issue" most surely is within your own Internet connection. I have been searching through my most recent 4 days of logs and cannot find a single occurrence of this error in my own network - overall, I don't think this is a Pi-hole issue. The code doing this has always been there, just the reporting to the user is new.

What may be of interest to further debug this is:

  1. Why does it only affect TCP queries? (usually, DNS is UDP)
  2. Is it always the same client making the query?
  3. Is it always for the same domain / query type?