AdguardTeam
"failed to sufficiently increase send buffer size" log error even though UDP buffer size was increased to ~2.5MB in /etc/sysctl.conf
Prerequisites
-
[X] I have checked the Wiki and Discussions and found no answer
-
[X] I have searched other issues and found no duplicates
-
[X] I want to report a bug and not ask a question
Operating system type
Linux, Other (please mention the version in the description)
CPU architecture
x86
Installation
GitHub releases or script from README
Setup
On a router, DHCP is handled by the router
AdGuard Home version
v0.107.31
Description
What did you do?
Upgraded container to v.107.31 (this started happening at the prior version v0.107.30). It does not happen in v0.107.29.
I had edited my /etc/sysctl.conf to add the line net.core.rmem_max=2500000 to allocate ~2.5MB of buffer
Expected result
No failed to sufficiently increase send buffer size error
Actual result
failed to sufficiently increase send buffer size error
Additional information
This started happening on all my adguardhome instances:
- Debian 11
- Ubuntu 20.04
- RPI Debian 11 64 bit
Rolling back to v0.107.29 fixes the issue
This is the full error message I receive on all my instances after upgrading to v0.107.30 . Still happening in v0.107.31 .
failed to sufficiently increase send buffer size (was: 208 kiB, wanted: 2048 kiB, got: 416 kiB). See https://github.com/quic-go/quic-go/wiki/UDP-Receive-Buffer-Size for details.
Is AdGuard Home running in a container or on the host system itself? Because in your issue template you've written that you've used the installation script as opposed to selecting the “Docker” option. If it's in a Docker container, there are known Docker issues with sysctl, and resolving those requires running the container in a more privileged manner.
I run AGH in Docker containers and I have set net.core.rmem_max=2500000 in the Docker host's /etc/sysctl.conf to mitigate the inability to change kernel sysctls from inside a Docker container.
This was working up until v0.107.29, and stopped working on v0.107.31, along with the other issue I reported on another post (#5902).
Currently I have one AGH container running v0.107.29 (to test out rolling back AGH version for #5902) and another one running the v0.107.31 and just upgraded to v0.107.32.
The one on v0.107.29 is running without UDP buffer complains, while the one on v0.107.32 is throwing the following message in the log:
failed to sufficiently increase send buffer size (was: 208 kiB, wanted: 2048 kiB, got: 416 kiB). See https://github.com/quic-go/quic-go/wiki/UDP-Receive-Buffer-Size for details.
Does this setting really matter for DNS queries over http/3? I can see how in general it's important for http/3 connections that move a lot of data, but DNS queries should be very small. Maybe it is important for AGH servers that see a lot of traffic and not an issue in a home setup.
Is AdGuard Home running in a container or on the host system itself? Because in your issue template you've written that you've used the installation script as opposed to selecting the “Docker” option. If it's in a Docker container, there are known Docker issues with
sysctl, and resolving those requires running the container in a more privileged manner.
Apologies, I am running in a docker container using a standard AGH docker compose stack.
Same problem here...
Just using
---
version: "2"
services:
adguardhome:
image: adguard/adguardhome
container_name: adguardhome
hostname: adguardhome
ports:
- 53:53/tcp
- 53:53/udp
- 784:784/udp
- 853:853/tcp
- 3000:3000/tcp
- 80:80/tcp
- 443:443/tcp
network_mode: "host"
restart: unless-stopped
#healthcheck:
#test: wget --no-verbose --tries=1 --spider http://localhost || exit 1
#interval: 120s
#retries: 5
#start_period: 20s
#timeout: 10s
labels:
- deunhealth.restart.on.unhealthy=true
volumes:
- /volume1/docker/adguardhome/work:/opt/adguardhome/work
- /volume1/docker/adguardhome/conf:/opt/adguardhome/conf
Add the NET_ADMIN capability to your container and things should work, as quic-go is using SO_RCVBUFFORCE / SO_SNDBUFFORCE as of v0.35.0
version: "2"
services:
adguardhome:
image: adguard/adguardhome
container_name: adguardhome
hostname: adguardhome
ports:
- 53:53/tcp
- 53:53/udp
- 784:784/udp
- 853:853/tcp
- 3000:3000/tcp
- 80:80/tcp
- 443:443/tcp
network_mode: "host"
cap_add:
- NET_ADMIN
restart: unless-stopped
#healthcheck:
#test: wget --no-verbose --tries=1 --spider http://localhost || exit 1
#interval: 120s
#retries: 5
#start_period: 20s
#timeout: 10s
labels:
- deunhealth.restart.on.unhealthy=true
volumes:
- /volume1/docker/adguardhome/work:/opt/adguardhome/work
- /volume1/docker/adguardhome/conf:/opt/adguardhome/conf
I had the same "problem" error message with Adguard in Docker on a macvlan network, but after adding NET_ADMIN to my docker-compose.yml as @otbutz suggested the error message is gone! Thanks!
Hello,
I have the same problem. I run AGH in a docker container under unRAID.
I tried running the container with the NET_ADMIN capabilities.
My host has :
net.core.rmem_max = 16777216
net.core.wmem_max = 16777216
Those parameters are unknown for the sysctl inside the container.
# sysctl -a | grep net.core.rmem
#
I get the error message with a dns-over-quic query, but I get an answer:
./dnslookup example.com quic://my.dns.server.address
dnslookup v1.10.0
2023/10/29 15:23:01 failed to sufficiently increase send buffer size (was: 208 kiB, wanted: 2048 kiB, got: 416 kiB). See https://github.com/quic-go/quic-go/wiki/UDP-Buffer-Sizes for details.
Server: quic://my.dns.server.address
dnslookup result (elapsed 43.258684ms):
;; opcode: QUERY, status: NOERROR, id: 26316
;; flags: qr rd ra; QUERY: 1, ANSWER: 1, AUTHORITY: 0, ADDITIONAL: 0
;; QUESTION SECTION:
;example.com. IN A
;; ANSWER SECTION:
example.com. 86400 IN A 93.184.216.34
But I also get the error with a http/3 query and it ends with a time out:
$ ./dnslookup example.com h3://my.dns.server.address/dns-query
dnslookup v1.10.0
2023/10/29 15:29:42 failed to sufficiently increase send buffer size (was: 208 kiB, wanted: 2048 kiB, got: 416 kiB). See https://github.com/quic-go/quic-go/wiki/UDP-Buffer-Sizes for details.
2023/10/29 15:29:47 [fatal] Cannot make the DNS request: requesting https://my.dns.server.address:443/dns-query: Get_0rtt "https://my.dns.server.address:443/dns-query?dns=AAABAAABAAAAAAAAB2V4YW1wbGUDY29tAAABAAE": timeout: no recent network activity
No error with https queries (or tls)
Same problem here:
Get this error on the Docker Container log when I setup quic servers to be used as DNS servers:
2025/05/06 15:13:23.945806 failed to sufficiently increase receive buffer size (was: 208 kiB, wanted: 7168 kiB, got: 416 kiB). See https://github.com/quic-go/quic-go/wiki/UDP-Buffer-Sizes for details.
When I go into the Docker container and try to modify this:
non-BSD
It is recommended to increase the maximum buffer size by running:
sysctl -w net.core.rmem_max=7500000 sysctl -w net.core.wmem_max=7500000
As per information found at https://github.com/quic-go/quic-go/wiki/UDP-Buffer-Sizes
I get this error:
/opt/adguardhome/work # sysctl -w net.core.rmem_max=7500000 sysctl: error: 'net.core.rmem_max' is an unknown key /opt/adguardhome/work # sysctl -w net.core.wmem_max=7500000 sysctl: error: 'net.core.wmem_max' is an unknown key /opt/adguardhome/work #
looks like Kernel isn't compiled with these.
When I go into the Docker container and try to modify this
You can't alter those parameters from within the container.
When I go into the Docker container and try to modify this
You can't alter those parameters from within the container.
It seems Adguard Home can't do it either. 2025/05/06 15:13:23.945806 failed to sufficiently increase receive buffer size (was: 208 kiB, wanted: 7168 kiB, got: 416 kiB). See https://github.com/quic-go/quic-go/wiki/UDP-Buffer-Sizes for details.