0

I have quite odd problem which i had for quite a time. I'm using powerful OVH dedicated server and i'm not sure what exactly is a root issue here.

I've migrated out of Cloudflare cdn, so this problem is quite annoying now and i need to finally tackle it. I have noticed that from time to time my server, randomly timeout connections. What i'm using:

  1. Nginx and hosting aprox ~15servers/domains via domains only.
  2. HHVM for some servers in Nginx, PHP-FPM for others
  3. Ubuntu 16.04.x

What i have noticed about issue:

  1. It does really happen randomly, sometimes timeout is longer 2-4min, sometimes few sec only. Sometimes it's limited to secluded cases, rarely, at least once per day for longer period for ~5-15s for more than 10 independent locations. It's possible that it won't happen for an hour, but it can happen many times in short period of time as well.
  2. At time of issue i still have active connection ssh connection to server and ICMP ping to server, both working fine - however httping fails at time of issue
  3. I haven't noticed any error logs in Nginx error.log tail -f error.log i thought that maybe proxy to hhvm/php-fpm fails, but there is no gateway issues whatsoever, neither php execution engines nor database is loaded heavily to cause any issue whatsoever.
  4. At time of issue, there are no logs appearing from "broken" ip which is connecting at time in nginx access.log tail -f access.log | grep "120.120.120.120"

I would really, highly appreciate any input to where look for to solve this as i'm out of ideas.

  • Question is slightly ambiguous. During this 2-4 minutes does the server time out all requests from an IP via http, or does one timeout take 2-4 minutes? Do you have any kind of firewall, before or on your instance, like iptables? Can you share your Nginx config and the server config? Could be you've set up rate limiting. How busy is the server? Does the request hit Nginx but not get to PHP / HHVM, or does it not reach Nginx? You'll need to check your logs for that info. – Tim May 27 '18 at 01:17
  • Timeout is executed by client limit i think. By 2-4 minutes i mean interval when all request to server from that ip time out. No firewall on server, nothing beyond default config with nginx, mariadb server and hhvm/php-fpm. Server is not busy whatsoever, it's processing ~50req per sec on average, Load average is below 0.2 for 16 cores. I do think that request do not reach nginx (?) - Nothing shown in access.log, but not sure when exactly log is created, whether as soon connection is established or when fully closed. –  May 27 '18 at 09:05

0 Answers0