Description of problem: When glusterd is down on one of the node in the cluster host should go to non operational which is not happening as of now. Host status is always shown as UP even when glusterd is down on that node. Version-Release number of selected component (if applicable): ovirt-engine-webadmin-portal-3.6.0-0.0.master.20150329172249.git660d494.el6.noarch How reproducible: Always Steps to Reproduce: 1. Add gluster nodes to ovirt 2. Now stop glusterd in one the node in the cluster. 3. Actual results: Ovirt still shows the node with up status and node does not go to non-operational. Expected results: When glusterd is down on the node, host status in UI should show it as "non-operational". Additional info:
Please note that glusterd is the management daemon and it's job is to look for other peers and distribute information across nodes. Once a client is connected to a gluster volume, it ends up talking directly to the brick processes - i.e glusterfsd processes. There is a glusterfsd process per volume. If you want to simulate one of the gluster nodes being unreachable from gluster perspective, shut down glusterd and kill the glusterfsd processes.
This happens only in the case when JSON RPC is used for communication. Exception handling is different in JSON RPC and currently its not handled properly for gluster related exception. As result, always 'JsonRpcInternalError"(Code: -32603) is returned from VDSM for any exception.
accidentally caught this with our script and moved it to "ON_QA" reverting back to how it was.
Verified and works fine with build ovirt-engine-3.6.0-0.0.master.20150517172245.git089e126.el6.noarch. When host is added to ovirt in json/xml rpc mode and if glusterd is down on that node ovirt moves the node to non operational.
This is an automated message. oVirt 3.5.3 has been released on June 15th 2015 and should include the fix for this BZ. Moving to closed current release.