2

We are frequently seeing (once or twice every day at least) "java.net.UnknownHostException" errors on some of our Google Kubernetes Engine pods. When it happens, these errors come for all external hosts our process on those pods are trying to reach, for instance:

  • java.net.UnknownHostException: datastore.googleapis.com
  • java.net.UnknownHostException accounts.google.com
  • com.google.gcloud.datastore.DatastoreException - I/O error

Anyone else has faced this issue? What could be the cause of this sudden loss of this connectivity on the pods? Off late (for last 2 weeks or so), we seem to be noticing this issue more frequently than before.

Thanks.

ahmet alp balkan
  • 42,679
  • 38
  • 138
  • 214
Roshan
  • 290
  • 3
  • 16
  • These sort of things are very hard to respond unless you have a good repro. Your description fits to DNS resolution error, which may go wrong at several different places. I would recommend running a container with bash command: `while true; do host datastore.googleapis.com; sleep 5; done` and checking the logs for errors. This may not reveal the issue reproducibly, because your OS will cache DNS records. – ahmet alp balkan Jun 26 '18 at 03:34
  • @AhmetAlpBalkan Thanks for the idea (and also for the edits). I'll try running this container. Anything else I can possibly try for this issue that shows up on select pods on select nodes / hosts? Even on the same node, some pod hits this issue and other pods don't, it seems. – Roshan Jun 26 '18 at 03:42

0 Answers0