-
Bug
-
Resolution: Done
-
Major
-
10.1.0.CR1
-
None
-
None
On a multi-interface Windows box
Even though the EAP balancer instance is bound to an IP address, 10.40.128.61 in this case:
<interfaces>
<interface name="management">
<inet-address value="10.40.128.61"/>
</interface>
<interface name="public">
<inet-address value="10.40.128.61"/>
</interface>
</interfaces>
, it advertises a hostname as its Manager-Address. The receiving worker node then might have troubles calling the balancer if it translates the hostname to a different IP address (interface on the same box...):
Balancer:
2016-04-28 18:00:00,232 DEBUG [io.undertow] (default I/O-2) UT005040: Gonna send payload:
HTTP/1.0 200 OK
Date: Thu, 28 Apr 2016 18:00:00 -0700
Sequence: 11
Digest: d356f44c4372f38eeb064e214b61b3e0
Server: 73a81e63-9938-42ce-bd17-92e2aeef3991
X-Manager-Address: atomic-0.yada.bla.bla.redhat.com:8847
X-Manager-Url: /
X-Manager-Protocol: https
X-Manager-Host: atomic-0.yada.bla.bla.redhat.com
Worker:
MODCLUSTER000043: Failed to send INFO command to atomic-0.yada.bla.bla.redhat.com/169.254.47.37:8847: Connection refused: connect
i.e. instead of using 10.40.128.61 the balancer is bound to, worker tries to call anotehr interface and fails.
Suggestion: If user explicitly uses IP address instead of a hostname, it should be either honoured or at least explicitly configurable for the Manager-Address attribute. IMHO.
- clones
-
-
- Verified
-