502 bad gateway, header size - node.js

Question

As our organization is using SSO for staff we are getting 502 bad gateway when users try to login with shibboleth.
The users who has more groups access, and try to login they are getting 502, but the users who has less access they are able to logged in.
The maximum header size with all the access is 32768.
We tried the --max-http-header-size 42768 in docker, how ever it was not helpful.
The users with normal access(less header size) is able to log in.
Our setup:
VM1 host the nginx as reverse proxy. The configuration is below.
VM2 host more than one docker.
server {
listen 80;
server_name **********;
proxy_buffering off;
proxy_set_header X-Real-IP $remote_addr;
proxy_set_header X-Forwarded-For $proxy_add_x_forwarded_for;
proxy_set_header Host $host;
client_body_timeout 60s;
client_header_timeout 60s;
keepalive_timeout 70s;
send_timeout 60s;
client_body_buffer_size 32k;
client_header_buffer_size 32k;
client_max_body_size 0;
large_client_header_buffers 4 32k;
access_log off;
error_log /data/nginx/logs/****_error.log warn;
location / {
proxy_pass http://******:8098;
}
}
Error log:
2019/09/25 10:25:38 [error] 20070#0: *123 upstream prematurely closed
connection while reading response header from upstream, client: ****,
server: ******, request: "GET /auth/shibboleth?redirect=L2FjY291bnQ=
HTTP/1.1", upstream: "http://******:8098/auth/shibboleth?redirect=L2FjY291bnQ=",
host: "*****", referrer:
"https://******/profile/SAML2/Redirect/SSO?execution=e1s2"
2019/09/25 10:25:50 [error] 20070#0: *125 upstream prematurely closed
connection while reading response header from upstream, client: ****,
server: *****, request: "GET / HTTP/1.1", upstream: "http://****:8098/",
host: "*****"
Docker setup
FROM node:8-alpine as intermediate
RUN apk add --no-cache git openssh alpine-sdk python2
RUN python2 -m ensurepip && \
rm -r /usr/lib/python*/ensurepip && \
pip install --upgrade pip setuptools && \
if [[ ! -e /usr/bin/python ]]; then ln -sf /usr/bin/python2
/usr/bin/python; fi
WORKDIR /usr/src/app
RUN touch config.js && mkdir config
COPY package*.json ./
RUN http_proxy="http://****:3128" https_proxy="http://****:3128" npm install
COPY . .
RUN rm -rf .private
FROM node:8-alpine
WORKDIR /usr/src/app
COPY --from=intermediate /usr/src/app /usr/src/app
EXPOSE 8080
CMD [ "node", "app.js", "-p 8080" ]

Answer 1

This is apparently common. The fix:
proxy_buffer_size 128k;
proxy_buffers 4 256k;
proxy_busy_buffers_size 256k;
See e.g.
Kubernetes nginx ingress controller returns 502 but only for AJAX/XmlHttpRequest requests - Stack Overflow
Fixing Nginx "upstream sent too big header" error when running an ingress controller in Kubernetes