My last encounter with oVirt ended with big W.T.F. So, Redhat <b>sells</b> th...
Tomasz Torcz
My last encounter with oVirt ended with big W.T.F. So, Redhat sells this stuff (as RHEV)‽ It sounds almost like a scam.
This whole business is in Java. For something as complicated as few tables holding machines' inventory. I know it was initially written in .NET, but RH could at least use Mono than convert it to Java.
Probably because of above, 768MB virtual machine wasn't enough to install. That is, before editing install script, which just checks if there is 2GB of RAM.
Next, there was this gem: Error: current locale (pl_PL.UTF-8) is not supported. supported locales are: en_US.UTF-8,en_US.utf-8,en_US.utf8. Uhm, right, in 2012.
Needs HTTP port. Great, this machine is dedicated for management node, so I'll put 80 here. ERROR: port is outside the range of 1024 - 65535
Finally, it needs FQDN. And a DNS server. What about my .local domain in lab?
Moving on to node installation, using 2.2.3-1.1 iso image. I grabbed some Optiplex 380 collecting dust in office and started install. It failed few times on rereadpt step; installation finished only after I wiped HDD clean.
Then some peculiarity: highest level category - "Datacenter" - is distinguished by storage type. You can select either local HDD on node or remote iSCSI or NFS. For added laughs, /data wasn't mounting automatically during node boot in either case. sigh
So I created test volume and tried to attach it to the node. No luck, "cannot connect to storage". Tried carefuly to follow every step from manual - doesn't work. Checked if other machine is able to access the iSCSI share - no problem here. So what's going on?
Digging through oVirt source and logs on the node I came across this: Thread-667::ERROR::2012-02-17 09:46:46,428::dispatcher::90::Storage.Dispatcher.Protect::(run) {'status': {'message': "Volume Group not big enough: ('VG size must be more than 10240 MiB',)", 'code': 504}}.
Oh right. And of course I made the test volume 8GiB. Let's grow it by two more gigs and try to connect. It works. Why isn't this documented and why the helpful error message is not shown to admin? sigh
Node image is generally well written. It's based on Fedora 16 with all the stateless goodness. It is so stateless that node losses association every reboot and needs to be re-added on management node.
A week later I have one virtual machine created on a node. I can turn it on, and off and on again. And that's all, because VNC connection to the console doesn't work. Spice connection "needs secure channel" and doesn't work, too. Neither from Firefox, Chromium nor standalone clients - vinagre and spicec.
And you can buy all this fail neatly packaged! Red Hat Enterprise Virtualization for Servers Starter Kit - Standard Subscription (1 year) is $2,994 USD.
I can't wait to see what +Piotr Baranowski will show at +OSEC Barcamp tomorrow :-)
This whole business is in Java. For something as complicated as few tables holding machines' inventory. I know it was initially written in .NET, but RH could at least use Mono than convert it to Java.
Probably because of above, 768MB virtual machine wasn't enough to install. That is, before editing install script, which just checks if there is 2GB of RAM.
Next, there was this gem: Error: current locale (pl_PL.UTF-8) is not supported. supported locales are: en_US.UTF-8,en_US.utf-8,en_US.utf8. Uhm, right, in 2012.
Needs HTTP port. Great, this machine is dedicated for management node, so I'll put 80 here. ERROR: port is outside the range of 1024 - 65535
Finally, it needs FQDN. And a DNS server. What about my .local domain in lab?
Moving on to node installation, using 2.2.3-1.1 iso image. I grabbed some Optiplex 380 collecting dust in office and started install. It failed few times on rereadpt step; installation finished only after I wiped HDD clean.
Then some peculiarity: highest level category - "Datacenter" - is distinguished by storage type. You can select either local HDD on node or remote iSCSI or NFS. For added laughs, /data wasn't mounting automatically during node boot in either case. sigh
So I created test volume and tried to attach it to the node. No luck, "cannot connect to storage". Tried carefuly to follow every step from manual - doesn't work. Checked if other machine is able to access the iSCSI share - no problem here. So what's going on?
Digging through oVirt source and logs on the node I came across this: Thread-667::ERROR::2012-02-17 09:46:46,428::dispatcher::90::Storage.Dispatcher.Protect::(run) {'status': {'message': "Volume Group not big enough: ('VG size must be more than 10240 MiB',)", 'code': 504}}.
Oh right. And of course I made the test volume 8GiB. Let's grow it by two more gigs and try to connect. It works. Why isn't this documented and why the helpful error message is not shown to admin? sigh
Node image is generally well written. It's based on Fedora 16 with all the stateless goodness. It is so stateless that node losses association every reboot and needs to be re-added on management node.
A week later I have one virtual machine created on a node. I can turn it on, and off and on again. And that's all, because VNC connection to the console doesn't work. Spice connection "needs secure channel" and doesn't work, too. Neither from Firefox, Chromium nor standalone clients - vinagre and spicec.
And you can buy all this fail neatly packaged! Red Hat Enterprise Virtualization for Servers Starter Kit - Standard Subscription (1 year) is $2,994 USD.
I can't wait to see what +Piotr Baranowski will show at +OSEC Barcamp tomorrow :-)
Shared with: Public, Piotr Baranowski, OSEC
+1'd by: Piotr Baranowski
Comments
Comments powered by Disqus
Not for the faint of heart :-)
BTW, my rant have bugs filled:
- 785728 - [installation] ovirt node installation fails
- 806349 - /data not mounted after reboot