NYC3 droplets having intermittent connectivity loss?

August 21, 2014 1.7k views

I created two droplets on NYC3. Both are CentOS 6.5 x86_64 and are still stock exactly as they are when the droplet is created. I have not upgraded any packages nor changed any configurations. I have not done anything to these droplets other than ping them or ssh into them and run top.

Both droplets have been experiencing several-minute periods of no ping response and are unresponsive. Support has so far not investigated this enough to confirm there is an issue. I should also note that I still have internet connectivity locally when these droplets start becoming unresponsive. I have also tried pinging them from different locations/service providers while they are down and have confirmed it's nothing on my end.

Is anyone else experiencing a similar issue on NYC3? If so, please speak up because so far support thinks I'm just some idiot who is breaking my droplets with bad iptables configs. My droplets are currently STOCK for the sole purpose of attempting to prove this issue is not my doing - thus I'm essentially wasting time I could be spending deploying our company's apps.

1 comment
  • Same problem, created a new droplet in NYC3 around an hour ago and still no internet access, doesn't even reach the default gateways.

    BTW did you configure it to have IPv6 and/or internal network as well?

2 Answers

From the point of view of support, they have to verify that you're not introducing bugs yourself with bad netfilter rules, because 99% of the time that's literally what's happening.

However, if you show them your configs like they ask, then they shouldn't insist further.

  • I submitted my configuration to them in a ticket yesterday and they dismissed the issue as a kernel version discrepancy (I was using the default kernel). After setting the kernel version to something else - same issue. I have also tried destroying and recreating the droplets.

    Now, as I said in the question - I am purposely running stock CentOS (installed via create droplet) without any changes to configuration or package updates in order to prove that this issue is not in the "99% bad netfilter rules" category.

    Do you even have the ability or expertise on staff to monitor your datacenters? I'm starting to suspect the answer is no, based on the responses in the tickets I've opened over this.

  • I'm also honestly quite baffled by these issues and why they take so long to resolve, but I can't help much more than this.

    They asked me if I was filtering ICMP but that's about it.


There does seem to have been an issue affecting some customers that we've since resolved. If you're still seeing any problems, please let us know in your ticket!

DigitalOcean Support

Have another answer? Share your knowledge.