Error 502 Bad gateway while writing Excel file with Gunicorn & Nginx

Question

Using Django

Whenever writing a small spreadsheet file, no problem. But when over 700 rows, keep getting 502 Bad gateway. Nginx error log shows that "upstream prematurely closed connection while reading response header from upstream". Based on this, it is indicated that the cause on backend side. However, Django error log file shows nothing.

Current settings in Nginx:

    http {
    include       /etc/nginx/mime.types;
    default_type  application/octet-stream;

    log_format  main  '$remote_addr - $remote_user [$time_local] "$request" '
                      '$status $body_bytes_sent "$http_referer" '
                      '"$http_user_agent" "$http_x_forwarded_for"';

    access_log  /var/log/nginx/access.log  main;

    sendfile        on;
    #tcp_nopush     on;

    #keepalive_timeout  0;
    keepalive_timeout  300s;
    proxy_read_timeout 300s;

    #gzip  on;

    # Load config files from the /etc/nginx/conf.d directory
    # The default server is in conf.d/default.conf

    include /etc/nginx/sites-enabled/*;
}

Gunicorn is being executed with Supervisord. Config lines are:

[program:gunicorn]
command=/home/enki10/.virtualenvs/license-project/bin/python /home/enki10/www/license-project/manage.py run_gunicorn -w 4 -k gevent -t 300
directory=/home/enki10/www/license-project
stdout_logfile=/var/log/gunicorn.log
stderr_logfile=/var/log/gunicorn_error.log
user=www-data
autostart=true
autorestart=true

Updated Here is the code related to creating spreadsheet file:

'''
This columns contains header in excel report
'''
columns = [
    (u"Origin", 10000),
    (u"Deal", 2000),
    (u"Customer", 10000),
    (u"Plate", 2500),
    (u"Serial", 6000),
    (u"Sold", 3000),
    (u"Clerk", 3000),
    (u"Received", 3000),
]

vehiclesales = VehicleSale.objects.filter(status__lt=AGENT_SUBMITTED_STATUS).order_by('origin', 'sold')

if vehiclesales:
    if format == EXPORT_EXCEL:
        response = HttpResponse(mimetype='application/ms-excel')
        response['Content-Disposition'] = 'attachment; filename=pending_cases.xls'
        workbook = xlwt.Workbook(encoding='utf-8')
        sheet = workbook.add_sheet("Pending cases")     

   row_num = 0
    font_style = xlwt.XFStyle()
    font_style.font.bold = True

    for col_num in xrange(len(columns)):
        sheet.write(row_num, col_num, columns[col_num][0], font_style)
        # set column width
        sheet.col(col_num).width = columns[col_num][1]

    font_style = xlwt.XFStyle()
    font_style.alignment.wrap = 1

    #here we go with filling actual data in sheet
    for item in vehiclesales:
        vehiclesale = vehiclesale_to_dict(item)
        row_num += 1
        row = [
            vehiclesale['origin'],
            vehiclesale['deal'],
            vehiclesale['customer'],
            vehiclesale['plate'],
            vehiclesale['vin'],
            vehiclesale['sold'],
            vehiclesale['clerk'],
            vehiclesale['received']
        ]

        for col_num in xrange(len(row)):
            if col_num == 5:
                font_style.num_format_str = 'MM/dd/yyyy'
            else:
                font_style.num_format_str = 'general'
            sheet.write(row_num, col_num, row[col_num], font_style)

    workbook.save(response)
    return response
else:
    no_report_template_data['report'] = 'Pending cases'
    return render_to_response(no_report_template_name, no_report_template_data)

Update finished.

Whether I include the timeout setting when calling gunicorn or not, it does make a difference, same behaviour: always error 502 at 30 seconds-mark

Also noticed, that if the quantity of rows is small (less than 400) but the quantity of columns are like 20, then it also fails. So it does not seem to fail because of rows, but how much data is processed.

Please help.