jbonofre opened a new pull request, #159:
URL: https://github.com/apache/karaf-cellar/pull/159

   ## Summary
   
   The shipped `etc/hazelcast.xml` enables multicast discovery with the 
well-known group
   name/password (`cellar`/`pass`) and no SSL, symmetric-encryption or 
socket-interceptor. In
   Hazelcast OSS, the group password is not an authentication mechanism - it 
isn't verified at
   join. Combined with `groups.cfg` shipping 
`default.bundle.whitelist.inbound=*` and
   `default.bundle.sync=cluster`, any host that can reach the multicast group 
can join the
   cluster and push a bundle-install cluster event that other nodes will 
execute.
   
   This PR does **not** change any default values (multicast, group 
name/password, or the
   `groups.cfg` whitelist) - that's a separate, more disruptive discussion. It 
only makes the
   risk impossible to miss and gives operators concrete hardening steps:
   
   * `HazelcastServiceFactory` now logs a loud, non-blocking `WARN` at startup 
if the group
     name/password are still the shipped defaults, and/or if multicast is 
enabled with no
     SSL/symmetric-encryption/socket-interceptor protecting it.
   * The shipped `hazelcast.xml` gets explanatory comments next to the 
`<group>` and
     `<multicast>` elements.
   * `hazelcast.adoc` no longer implies the group password is a real 
credential, and gains a
     new "Security considerations" section with hardening recommendations 
(network isolation,
     tcp-ip with an explicit member list instead of multicast, 
SSL/symmetric-encryption).
   
   ## Test plan
   
   - [x] `mvn -pl hazelcast -am test` - all tests pass, including a new
     `HazelcastServiceFactorySecurityTest` covering both detection conditions 
(default group,
     unprotected multicast) and their negative cases, plus a regression guard 
that asserts the
     shipped `assembly/src/main/resources/hazelcast.xml` is still detected as 
insecure today.
   - [x] Confirmed the existing `HazelcastServiceFactoryTest`, which spins up a 
real
     `HazelcastInstance`, still passes with the new check wired into 
`buildInstance()`.
   - [x] Manually reviewed the rendered `hazelcast.adoc` changes for correct 
AsciiDoc syntax.


-- 
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.

To unsubscribe, e-mail: [email protected]

For queries about this service, please contact Infrastructure at:
[email protected]

Reply via email to