[Users] Discussion... work appears to be needed in areas of recovery and cleanup
Hi... I've been working with both RHEV and Ovirt and can see distinct changes and improvements with regards to how Ovirt is emerging and changing. One area that still disturbs me is in the area of recovery and cleanup, specifically around the ovirt-engine (rhevm) node. I've observed a few things about the project: - backup of a engine db seems to be a fairly manual process with many steps. There are many files listed that need to be backed up manually as listed in the RHEVM admin manual. pg_dump seems to automate the dumping of the db, but other config files are left up to the admin. - restoration appears to be equally complex - engine-cleanup (rhevm-cleanup) doesn't always cleanup - there appear to be instances where the only way to get a solid base is reinstallation. I've removed and reinstalled components so many times it is now second nature. ;) I'm curious about: - Are there technical issues (beyond the code not being written yet) that prevent backup and restore procedures from working correctly to save and restore an engine environment? I would expect that a test for that operation would be to setup an environment, take a snapshot, and restore the whole thing to another management node. - Are all the config files properly identified? The RHEVM Admin manual has an extensive list but is it exhaustive? - What technically prevents engine-cleanup from really *cleaning* a system? My expectation would be that I should be able to run engine-cleanup and completely wipe out the setup. All databases, certs, config files and the like returned to their original, pristine state. Perhaps another script - engine-restore-defaults? - needs to go beyond where engine-cleanup goes today? As a support professional, these are areas that I think warrant investment and I wonder what is gating the development here. Thanks, Rick -- +------+ | Rick Beldin| Hewlett-Packard Company| | email: rick.bel...@hp.com | | |TCE&Q - Linux ERT | Windward Parkway West | || Alpharetta, GA 30004 | +--+ ___ Users mailing list Users@ovirt.org http://lists.ovirt.org/mailman/listinfo/users
Re: [Users] Adding Iscsi storage to Ovirt All-in-one config
On 01/13/2013 02:09 AM, Itamar Heim wrote: > On 01/13/2013 01:25 AM, Tom Brown wrote: >>> you still can't mix different types of storage (local/iscsi) in same DC, so >>> you need to add another DC for the hosts which will use the iscsi storage. >> Thanks for pointing this out. I missed a major conceptual point here. Does this mean you might as well not have any local storage on a node beyond that to support the hypervisor? How do you move the virtual disks already deployed on one type of storage to another? While not common, I could expect this activity to occur as organizations grow their environment. Thanks, Rick ___ Users mailing list Users@ovirt.org http://lists.ovirt.org/mailman/listinfo/users
[Users] Adding Iscsi storage to Ovirt All-in-one config
I set up an ovirt 3.1 all-in-one config with Fedora and added one node. Created virtual machines and so forth. All seems to work well. I then decided to add iscsi and can verify that both the ovirt node and the all-in-one node can 'see' the storage even though it is uninitialized and does not have a partition table: # parted /dev/mapper/1IET_00010001 p Error: /dev/mapper/1IET_00010001: unrecognised disk label Model: Linux device-mapper (multipath) (dm) Disk /dev/mapper/1IET_00010001: 52.4GB Sector size (logical/physical): 512B/512B Partition Table: unknown Disk Flags: >From the OS on both the all-in-one host and the node, I can do 'normal' iscsi things and storage things without error. I created a new iscsi domain off the Default datacenter and it comes up although it says it is unattached.If I attempt to attach it I get: There are No Data Centers to which the Storage Domain can be attached. I'm confused. What am I missing out of my config to get iscsi added after the fact? Or is this something you can only do at installation time? It seems like this is one of those many things that need to be done in a certain order to get it to work. Thanks, Rick ___ Users mailing list Users@ovirt.org http://lists.ovirt.org/mailman/listinfo/users
Re: [Users] What do you want to see in oVirt next?
On Tue 08 Jan 2013 04:41:01 AM EST, Itamar Heim wrote: > On 01/07/2013 06:11 PM, Rick Beldin wrote: >> - cleaner work-flow in creating and associating storage, especially NFS >> Some of this is no doubt my newbie-ness to ovirt. Most of my comments below have to do with the manager. I started playing with RHEV 3.1 since I am responsible for delivering support on it but just yesterday started playing with Fedora 17 and all-in-one for ease of setup. I will try and document the specifics, but the things I recall were things like: - inability to setup NFS storage on the cluster until I had added a host. I think I missed some key concept here, but my feeling was that I would/could/should setup the manager (engine) first and then add virtual hosts. Instead, there seems to be a procedural step. - along those lines the UI could do more to guide a user, a first-time wizard that would guide you through cluster and datacenter setups that are independent of the hosts. Guide Me is a good start, but perhaps it needs some expansion. >> - better error reporting from engine back to admin user during admin >> operations. > - here's an example. Just installed AIO on Fedora 17. After going through everything, it says 'Install Failed'. The Event entry has a correlation id, which can be used to figure out what went wrong. (I guess). It seems like there could be more information provided to the admin on what to do next aside install again? Tooltip on what to do with correlation id? Oddly, I can't right click on Event text to bring up copy/paste menu with Chrome or FF. Ctrl-c worked with Chrome but not FF: Host local_host installation failed. The required action is taking longer than allowed by configuration. Verify host networking and storage settings.. That gives you a clue, but then what? A handy pointer to the log file would be good. - Eons ago, in another product space, a complex product with hooks into networking and whatnot had a tiny diagnostic script that would attempt to emulate the requirements of the product. Perhaps that is needed here - something outside of the framework that could be a checklist of ports not open, permissions, etc. Things that an admin needs to address outside of engine. - A menu item that would call the sosreport-like thing to collect data would aid in supportability. - A meta-page with pointers/contents to all logs, databases and so forth would be helpful. There is a lot of content scattered over the disk. - A meta-page with a listing of critical processes. Some of these things are doubtless outside the scope of abstraction that is currently there, but the manager is the connection between the admin and the hosts. As a critical resource the admin will likely not only be the one administering to guests, but to the engine as well. Making that person's life easier will aid in acceptance and reduce long-term support costs. My 2cents... Rick ___ Users mailing list Users@ovirt.org http://lists.ovirt.org/mailman/listinfo/users
Re: [Users] What do you want to see in oVirt next?
Several topics come to mind: - cleaner work-flow in creating and associating storage, especially NFS - better error reporting from engine back to admin user during admin operations. - the ability to have users access consoles of virtual machines whereever they may be. I think it should be expected that limited access data centers will be everywhere, from complete NAT'd environments to some with external firewalls that limit some ports. This should be part of engine and hypervisor installation, and transparent to the end user. ___ Users mailing list Users@ovirt.org http://lists.ovirt.org/mailman/listinfo/users
