Didi, Both source and destination machines (assuming you are referring to the physical host source and virtual host destination of the migration) are running ovirt-engine 3.5.2.1-1. However, the "source" is on CentOS 6.6, whereas the "destination" is on CentOS 7.1. The source is running postgresql(/-server) 8.4.20, Java 1.7.0_79, httpd 2.2.15. Destination is running postgresql(/-server) 9.2.10, Java 1.7.0.79, jboss-httpd 2.4.6. I'm not sure which over package version is relevant. The failure occurs after engine-setup, and the web UI doesn't load, and the host reports the engine is not up (I'm not in front of the host, so I don't have the exact language). Our current setup has 2 physical hosts, ~15 active VMs. Each host has 64GB RAM. Even with all the VM on one host, the utilization is less than 50%. We have another physical storage server, that provides the NFS storage backend to the hosts and VMs. Prior to attempting the hosted-engine migration, I migrated all the VMs to host #1, where the engine was originally installed. I then removed host #2 from the host list via the UI, and performed hosted-engine --deploy on host #2. A few other quirks I encountered: 1. I originally wanted to do "hosted-engine --deploy" on host #1, but that it was unsuccessful due to it complaining the storage connection was busy (or something). 2. I then tried to do "hosted-engine --deploy" on host #2, using the same storage connection path as what I used for host #1 originally. The setup process asked if it was an "additional host", I said no, but later it cannot activate the storage connection. 3. Lastly I did "hosted-engine --deploy" on host #2, but pointing to a different storage connection path, still on the same storage server. This was successful until the I encountered the failure in question. Thanks, Haven On Tue, May 26, 2015 at 2:20 AM, Yedidyah Bar David <didi@redhat.com> wrote:
----- Original Message -----
From: "Haoyang Haven Liu" <haven.liu@ucla.edu> To: users@ovirt.org Sent: Tuesday, May 26, 2015 8:34:00 AM Subject: [ovirt-users] hosted engine does not start correctly (CentOS 7.1, overt-engine 3.5.2)
Hello,
I tried to follow < http://www.ovirt.org/Migrate_to_Hosted_Engine > to migrate an existing ovirt environment. I managed to complete the “engine-backup —mode=restore…” step, and the following “engine-setup” step appear to have completed. However, the ovirt-engine service was not started successfully has errors:
[root@ovirt ~]# systemctl status ovirt-engine ovirt-engine.service - oVirt Engine Loaded: loaded (/usr/lib/systemd/system/ovirt-engine.service; enabled) Active: failed (Result: exit-code) since Mon 2015-05-25 22:26:12 PDT; 43s ago Main PID: 15931 (code=exited, status=1/FAILURE)
May 25 22:26:11 ovirt.polysep.ucla.edu systemd[1]: Started oVirt Engine. May 25 22:26:12 ovirt.polysep.ucla.edu ovirt-engine.py[15931]: 2015-05-25 22:26:12,269 ovirt-engine: ERROR run:532 Error: process termina...code 1 May 25 22:26:12 ovirt.polysep.ucla.edu systemd[1]: ovirt-engine.service: main process exited, code=exited, status=1/FAILURE May 25 22:26:12 ovirt.polysep.ucla.edu systemd[1]: Unit ovirt-engine.service entered failed state. Hint: Some lines were ellipsized, use -l to show in full.
Additionally, /var/log/ovirt-engine/console.log complains “cannot allocate memory”. The VM engine has 4GB of memory, but ovirt-engine engine seems to want ~11GB of memory, is that right?
[root@ovirt ~]# cat /var/log/ovirt-engine/console.log OpenJDK 64-Bit Server VM warning: INFO: os::commit_memory(0x0000000412180000, 11250696192, 0) failed; error='Cannot allocate memory' (errno=12) # # There is insufficient memory for the Java Runtime Environment to continue. # Native memory allocation (malloc) failed to allocate 11250696192 bytes for committing reserved memory. # An error report file with more information is saved as: # /tmp/jvm-15974/hs_error.log [root@ovirt ~]# free -m total used free shared buff/cache available Mem: 3791 123 3175 28 492 3428 Swap: 2559 0 2559 [root@ovirt ~]#
The hs_error.log is attached.
Any help is appreciated.
Can you please compare versions of all installed packages between the source and destination machines?
I assume the failure above is right after engine-setup, right? Before continuing the interaction on hosted-engine side?
Did you try to restart the engine (systemctl restart ovirt-engine)? Did it fail the same way?
How much memory do you have on the source machine? And how much does the engine actually use there?
How large is your system (hosts, VMs etc)?
Best, -- Didi