Filelist >500 files leads to close connection

[donkeyDau]

I'm experiencing a similar issue to #55 When sending a request with more than 500 (don't know the exact limit but something around this) I see:

2020/11/18 15:16:14 [alert] 12#12: *41 sendfile() failed (9: Bad file descriptor) while sending request to upstream, client: 192.168.79.196, server: , request: "POST /download/output/ckhbq1z1v000f08gie9weabiu HTTP/1.1", subrequest: "/s3/output/ckhbq1z1v000f08gie9weabiu/out/1002610", upstream: "http://10.96.147.248:80/output/ckhbq1z1v000f08gie9weabiu/out/1002610", host: "download.xxxx.com", referrer: "https://xxxx.com/output/export"
2020/11/18 15:16:14 [error] 12#12: *41 mod_zip: a subrequest returned 502, aborting... while sending to client, client: 192.168.79.196, server: , request: "POST /download/output/ckhbq1z1v000f08gie9weabiu HTTP/1.1", subrequest: "/s3/output/ckhbq1z1v000f08gie9weabiu/out/1002610", upstream: "http://10.96.147.248:80/output/ckhbq1z1v000f08gie9weabiu/out/1002610", host: "download.xxxx.com", referrer: "https://xxxx.com/output/export"
192.168.79.196 - - [18/Nov/2020:15:16:14 +0000] "POST /download/output/ckhbq1z1v000f08gie9weabiu HTTP/1.1" 200 0 "https://xxxx.com" "Mozilla/5.0 (X11; Linux x86_64; rv:83.0) Gecko/20100101 Firefox/83.0"

The file 1002610 is the first in the list.

I'm using nginx build with this docker file:

FROM centos:8

EXPOSE 80

ENV NGINX_VERSION=1.16.1

# install system packages
RUN yum -y install epel-release \
    && yum -y install gcc git make openssl-devel pcre-devel zlib-devel \
    supervisor \
    && yum -y update \
    && yum -y clean all

# install nginx
RUN curl -LO http://nginx.org/download/nginx-${NGINX_VERSION}.tar.gz \
    && tar zxf nginx-${NGINX_VERSION}.tar.gz \
    && cd nginx-${NGINX_VERSION} \
    && git clone https://github.com/evanmiller/mod_zip.git \
    && ./configure \
    --prefix=/usr \
    --conf-path=/etc/nginx/nginx.conf \
    --sbin-path=/usr/sbin \
    --http-log-path=/dev/stdout \
    --error-log-path=/dev/stdout \
    --lock-path=/var/lock/nginx.lock \
    --pid-path=/run/nginx.pid \
    --http-client-body-temp-path=/var/lib/nginx/body \
    --http-fastcgi-temp-path=/var/lib/nginx/fastcgi \
    --http-proxy-temp-path=/var/lib/nginx/proxy \
    --http-scgi-temp-path=/var/lib/nginx/scgi \
    --http-uwsgi-temp-path=/var/lib/nginx/uwsgi \
    --with-pcre-jit \
    --with-ipv6 \
    --with-file-aio \
    --with-threads \
    --with-http_ssl_module \
    --with-http_realip_module \
    --with-http_addition_module \
    --with-http_gzip_static_module \
    --with-http_gunzip_module \
    --with-http_sub_module \
    --add-module=mod_zip \
    && make install  \
    && cd .. \
    && rm -f nginx-${NGINX_VERSION}.tar.gz \
    && rm -rf nginx-${NGINX_VERSION} \
    && mkdir -p /var/lib/nginx/body


RUN mkdir -p /var/www/html

COPY start.sh /
COPY nginx.conf /etc/nginx/

CMD [ "/start.sh", "-c", "/etc/nginx/nginx.conf" ]

A fresh build 1 hour ago didn't change anything. The S3 backend is Minio.

I'm glad to provide more logs/details if needed.

[donkeyDau]

Further investigations led to a slightly different picture. I have two types of requests: one is carrying a POST body, the other not. Without the POST body I see no limitations. With POST body the previously stated "limit" kicks in and leads to the error. The POST body size should be something around 40-50kb when this error occurs. I'm stuck here. Has anyone a suggestion?