Created attachment 952052 [details] Hosted Engine Setup log on first node Description of problem: On rhel 6.6, after having installed on node1 hosted-engine and having complete the installation of the RHEVM VM on an NFS Share, when we repeat on node2 the command: hosted-engine --deploy and insert the NFS storage domain info, the already installed manager is not recognized and I am only offered to do a new manager installation: Please specify the storage you would like to use (nfs3, nfs4)[nfs3]: nfs4 Please specify the full shared storage connection path to use (example: host:/path): virt1.gcio.unicredit.eu:/ [ INFO ] Installing on first host Please provide storage domain name. [hosted_storage]: Version-Release number of selected component (if applicable): RHEL 6.6 RHEV 3.4 ovirt-host-deploy-1.2.3-1.el6ev.noarch ovirt-hosted-engine-ha-1.1.5-1.el6ev.noarch ovirt-hosted-engine-setup-1.1.5-1.el6ev.noarch How reproducible: * OS Setup install rhel 6.6 on 2 nodes install ovirt-hosted-engine-setup from rhev 3.4 on both * Create an NFS4 export # mkdir -p /nfs/rhev-manager # chown 36:36 /nfs/rhev-manager Explicitly set Domain in /etc/idmapd.conf export it as: /nfs/rhev-manager *(rw,no_root_squash,fsid=0,sync) restart NFS * test mount from both nodes: # mount -t nfs4 nfshost:/ /mnt result: both nodes ok * RHEV Setup on node1 # hosted-engine --depoly pass the nfs4 share above. Install the rhevm on the VM created by the setup. Complete the procedure without any problem. results: - ssh on RHEVM is ok - admin console ok - status hosts ok on node2 # hosted-engine --depoly ... Please specify the storage you would like to use (nfs3, nfs4)[nfs3]: nfs4 Please specify the full shared storage connection path to use (example: host:/path): nfshost:/ [ INFO ] Installing on first host <-- NO! Please provide storage domain name. [hosted_storage]: Actual results: The RHEVM installed on NFS4 is not detected. Expected results: hosted-engine should detect the already installed RHEVM VM on the NFS4 share.
Created attachment 952053 [details] Hosted Engine Setup log on second node
Ok, found it: for some reason it's removing the trailing '/' from remotePath attribute and so than it doesn't match all the storage pool attributes. So it thinks that is a new install and so it assumes that it's the first host. From the log of the second host: 2014-10-27 18:07:39 DEBUG otopi.plugins.ovirt_hosted_engine_setup.storage.storage storage._getStorageDomainInfo:432 {'status': {'message': 'OK', 'code': 0}, 'info': {'uuid': '7ec72ae9-44bd-408e-93f4-1fab1c881f69', 'version': '3', 'role': 'Master', 'remotePath': 'virt1.gcio.unicredit.eu:', 'type': 'NFS', 'class': 'Data', 'pool': ['4e9971cf-0c2b-4bd8-adb1-54a415fa8bc4'], 'name': 'hosted_storage'}} I'm investigating it.
It's fully reproducible also on oVirt 3.5.0 with vdsm 4.16.7-1. The problem is that VDSM is removing the trailing '/' from NFS remotePath attribute also when the user is simply trying to connect to the NFS global root of NFSv4 resulting in an invalid path identifier. [root@thec653 ~]# vdsClient -s 0 getStorageDomainsList 21fcaaec-ed51-4a17-adcb-eb6b5a087eba [root@thec653 ~]# vdsClient -s 0 getStorageDomainInfo 21fcaaec-ed51-4a17-adcb-eb6b5a087eba uuid = 21fcaaec-ed51-4a17-adcb-eb6b5a087eba version = 3 role = Master remotePath = 192.168.1.115: type = NFS class = Data pool = ['73da157b-322e-4b00-b4cf-ee3aeac0962d'] name = hosted_storage
As a temporary workaround, have you tried to give use "nfshost://" ?
Unfortunately the customer has moved the rhevm from a NFS4 to an NFS3 storage for the hosted engine, so I'm not able anymore to do any kind of test of this side.
For now as a workaround please do not use mount that is on "/". We will try to res love this for 3.6.
Reproduced internally by QA. We will continue tracking on bug 1215967. *** This bug has been marked as a duplicate of bug 1215967 ***