unable to resolve external dns names inside the warden

56 views
Skip to first unread message

mira...@gmail.com

unread,
Sep 18, 2014, 7:19:04 AM9/18/14
to vcap...@cloudfoundry.org
Hi,

I deployed cf on openstack+microbosh(recursor=8.8.8.8)+cf183

Then I deployed java application, on staging phase it couldn't resolve "github.com" name so I created offline buildpack as workaround and deployed the app successfully.

My app is working with external resources, but they are not available by dns name, only by ip.

I have 2 DEA runners, I checked /etc/resolv.conf:
nameserver 10.0.47.3
domain novalocal
search novalocal

then I ran this command:
host -a github.com 10.0.47.3
Trying "github.com"
;; Truncated, retrying in TCP mode.
Trying "github.com"
Using domain server:
Name: 10.0.47.3
Address: 10.0.47.3#53
Aliases:

;; ->>HEADER<<- opcode: QUERY, status: NOERROR, id: 51675
;; flags: qr rd ra; QUERY: 1, ANSWER: 14, AUTHORITY: 0, ADDITIONAL: 0

;; QUESTION SECTION:
;github.com. IN ANY

;; ANSWER SECTION:
github.com. 3599 IN SOA ns1.p16.dynect.net. hostmaster.github.com. 1411038649 3600 600 604800 60
github.com. 29 IN A 192.30.252.129
github.com. 29 IN A 192.30.252.130
github.com. 29 IN A 192.30.252.131
github.com. 29 IN A 192.30.252.128
github.com. 299 IN TXT "v=spf1 ip4:192.30.252.0/22 include:_spf.google.com include:esp.github.com include:cmail1.com include:mail.zendesk.com include:auth.madmimi.com ~all"
github.com. 299 IN SSHFP 2 1 7491973E5F8B39D5327CD4E08BC81B05F7710B49
github.com. 299 IN SSHFP 1 1 BF6B6825D2977C511A475BBEFB88AAD54A92AC73

Vms state:
+------------------------------------+---------+---------------+-------------+
| Job/index                          | State   | Resource Pool | IPs         |
+------------------------------------+---------+---------------+-------------+
| api_worker_z1/0                    | running | small_z1      | 10.0.47.6   |
| api_z1/0                           | running | large_z1      | 10.0.47.18  |
| clock_global/0                     | running | medium_z1     | 10.0.47.9   |
| etcd_z1/0                          | running | medium_z1     | 10.0.47.16  |
| ha_proxy_z1/0                      | running | router_z1     | 10.0.47.23  |
|                                    |         |               |**** |
| hm9000_z1/0                        | running | medium_z1     | 10.0.47.8   |
| loggregator_trafficcontroller_z1/0 | running | small_z1      | 10.0.47.4   |
| loggregator_z1/0                   | running | medium_z1     | 10.0.47.7   |
| login_z1/0                         | running | medium_z1     | 10.0.47.10  |
| logs_z1/0                          | running | medium_z1     | 10.0.47.15  |
| nats_z1/0                          | running | medium_z1     | 10.0.47.14  |
| nfs_z1/0                           | running | medium_z1     | 10.0.47.11  |
| postgres_z1/0                      | running | medium_z1     | 10.0.47.13  |
| router_z1/0                        | running | router_z1     | 10.0.47.22  |
| runner_z1/0                        | running | runner_z1     | 10.0.47.20  |
| runner_z1/1                        | running | runner_z1     | 10.0.47.19  |
| stats_z1/0                         | running | small_z1      | 10.0.47.5   |
| uaa_z1/0                           | running | medium_z1     | 10.0.47.12  |
+------------------------------------+---------+---------------+-------------+

Actually I'm running out ideas where to dig in

sparam...@pivotal.io

unread,
Sep 18, 2014, 11:49:09 AM9/18/14
to vcap...@cloudfoundry.org, mira...@gmail.com
Can you try running just curl command to see if it can connect? Quite possible there is timeout happening before the name could be resolved.

I could successfully run curl with verbose option against https://www.github.com as well as www.nytimes.com inside DEA of bosh-lite but it took real long.

I suspect the same thing happening with the buildpack download.
Change the TIMEOUT from 10 to 20 seconds inside the <java-buildpack>/lib/java_buildpack/util/cache/download_cache.rb file and see if the app pushes succeed.

-Sabha

Ferran Rodenas

unread,
Sep 18, 2014, 2:07:41 PM9/18/14
to vcap...@cloudfoundry.org, mira...@gmail.com
Check also you CF security groups to be sure you explicitly allow DNS queries from inside the DEAs. As an example, the AWS templates creates a global CF security group for DNS: https://github.com/cloudfoundry/cf-release/blob/master/templates/cf-properties.yml#L22

- Ferdy

--
You received this message because you are subscribed to the Google Groups "Cloud Foundry Developers" group.
To view this discussion on the web visit https://groups.google.com/a/cloudfoundry.org/d/msgid/vcap-dev/16eb062f-0c54-4206-8d29-bd52e831bec2%40cloudfoundry.org.

To unsubscribe from this group and stop receiving emails from it, send an email to vcap-dev+u...@cloudfoundry.org.

mira...@gmail.com

unread,
Sep 19, 2014, 4:34:19 AM9/19/14
to vcap...@cloudfoundry.org, mira...@gmail.com


четверг, 18 сентября 2014 г., 21:07:41 UTC+3 пользователь ferdy написал:
Thanks for reply this is exactly what was wrong, there were cf security groups but they haven't been assigned to staging and running states, I assigned them and now staging phase is resolving github.com etc and app from warden resolves external dns names.

Thanks a lot ! 
I just forgot about cf security groups. 
Reply all
Reply to author
Forward
0 new messages