Re: [ovirt-users] Help understanding Gluster in oVirt
I have setup an environment with 8 glustefs nodes in distributed+replicated mode. Instead of CTDB, I'm using keepalived. It works pretty well. On 02/04/2015 04:14 PM, Donny Davis wrote: I have two nodes running the hosted engine and gluster setup for replicate between them. Mounting to localhost seems to be working so far. I have live migrated the engine, pulled the plug on a running node, and rebooted each node to ensure it all works properly. I have had few issues with this setup. I didn’t like ctdb because it adds another layer of complexity that is unnecessary for MY setup. Node1 gluster drive on sdb Node2 gluster drive on sdb Both replicate and are the same make and model drive. ovirt-ha running on both for hosted-engine This setup is for hosted engine only, my actual storage for VM’s and what not do not run in this fashion. Hope This helps Donny D *From:*users-boun...@ovirt.org [mailto:users-boun...@ovirt.org] *On Behalf Of *Groten, Ryan *Sent:* Wednesday, February 4, 2015 10:11 AM *To:* Sahina Bose; users@ovirt.org *Subject:* Re: [ovirt-users] Help understanding Gluster in oVirt Nope in fact I followed the guide and found CTDB works quite well. I am just trying to figure out the benefit because that would be another component to consider in the architecture. *From:*Sahina Bose [mailto:sab...@redhat.com] *Sent:* Tuesday, February 03, 2015 4:09 AM *To:* Groten, Ryan; users@ovirt.org mailto:users@ovirt.org *Subject:* Re: [ovirt-users] Help understanding Gluster in oVirt On 01/28/2015 08:59 AM, Groten, Ryan wrote: I was planning on making a Gluster Data domain to test, and found some great information on this page: http://community.redhat.com/blog/2014/05/ovirt-3-4-glusterized/ The article the author uses the CTDB service for VIP failover. Is it possible/recommended to not do this, and just create a gluster volume on all the hosts in a cluster, then create the Gluster data domain as localhost:gluster_vol? Theoretically, it should work - if you make sure that you have a replica 3 gluster volume spread across 3 nodes, and these 3 nodes are your compute nodes as well - you should be fine without CTDB setup for failover and mounting as localhost. But I've not tried this to recommend it. Maybe if others have tried it, they can chime in? Btw, is there any reason you do not want to set up CTDB? Thanks, Ryan ThTh ___ Users mailing list Users@ovirt.org mailto:Users@ovirt.org http://lists.ovirt.org/mailman/listinfo/users ___ Users mailing list Users@ovirt.org http://lists.ovirt.org/mailman/listinfo/users ___ Users mailing list Users@ovirt.org http://lists.ovirt.org/mailman/listinfo/users
Re: [ovirt-users] Help understanding Gluster in oVirt
Nope in fact I followed the guide and found CTDB works quite well. I am just trying to figure out the benefit because that would be another component to consider in the architecture. From: Sahina Bose [mailto:sab...@redhat.com] Sent: Tuesday, February 03, 2015 4:09 AM To: Groten, Ryan; users@ovirt.org Subject: Re: [ovirt-users] Help understanding Gluster in oVirt On 01/28/2015 08:59 AM, Groten, Ryan wrote: I was planning on making a Gluster Data domain to test, and found some great information on this page: http://community.redhat.com/blog/2014/05/ovirt-3-4-glusterized/ The article the author uses the CTDB service for VIP failover. Is it possible/recommended to not do this, and just create a gluster volume on all the hosts in a cluster, then create the Gluster data domain as localhost:gluster_vol? Theoretically, it should work - if you make sure that you have a replica 3 gluster volume spread across 3 nodes, and these 3 nodes are your compute nodes as well - you should be fine without CTDB setup for failover and mounting as localhost. But I've not tried this to recommend it. Maybe if others have tried it, they can chime in? Btw, is there any reason you do not want to set up CTDB? Thanks, Ryan ThTh ___ Users mailing list Users@ovirt.orgmailto:Users@ovirt.org http://lists.ovirt.org/mailman/listinfo/users ___ Users mailing list Users@ovirt.org http://lists.ovirt.org/mailman/listinfo/users
Re: [ovirt-users] Help understanding Gluster in oVirt
I have two nodes running the hosted engine and gluster setup for replicate between them. Mounting to localhost seems to be working so far. I have live migrated the engine, pulled the plug on a running node, and rebooted each node to ensure it all works properly. I have had few issues with this setup. I didn't like ctdb because it adds another layer of complexity that is unnecessary for MY setup. Node1 gluster drive on sdb Node2 gluster drive on sdb Both replicate and are the same make and model drive. ovirt-ha running on both for hosted-engine This setup is for hosted engine only, my actual storage for VM's and what not do not run in this fashion. Hope This helps Donny D From: users-boun...@ovirt.org [mailto:users-boun...@ovirt.org] On Behalf Of Groten, Ryan Sent: Wednesday, February 4, 2015 10:11 AM To: Sahina Bose; users@ovirt.org Subject: Re: [ovirt-users] Help understanding Gluster in oVirt Nope in fact I followed the guide and found CTDB works quite well. I am just trying to figure out the benefit because that would be another component to consider in the architecture. From: Sahina Bose [mailto:sab...@redhat.com] Sent: Tuesday, February 03, 2015 4:09 AM To: Groten, Ryan; users@ovirt.org Subject: Re: [ovirt-users] Help understanding Gluster in oVirt On 01/28/2015 08:59 AM, Groten, Ryan wrote: I was planning on making a Gluster Data domain to test, and found some great information on this page: http://community.redhat.com/blog/2014/05/ovirt-3-4-glusterized/ The article the author uses the CTDB service for VIP failover. Is it possible/recommended to not do this, and just create a gluster volume on all the hosts in a cluster, then create the Gluster data domain as localhost:gluster_vol? Theoretically, it should work - if you make sure that you have a replica 3 gluster volume spread across 3 nodes, and these 3 nodes are your compute nodes as well - you should be fine without CTDB setup for failover and mounting as localhost. But I've not tried this to recommend it. Maybe if others have tried it, they can chime in? Btw, is there any reason you do not want to set up CTDB? Thanks, Ryan ThTh ___ Users mailing list Users@ovirt.org http://lists.ovirt.org/mailman/listinfo/users ___ Users mailing list Users@ovirt.org http://lists.ovirt.org/mailman/listinfo/users
Re: [ovirt-users] Help understanding Gluster in oVirt
- Original Message - From: Ryan Groten ryan.gro...@stantec.com To: users@ovirt.org Sent: Tuesday, January 27, 2015 7:29:55 PM Subject: [ovirt-users] Help understanding Gluster in oVirt I was planning on making a Gluster Data domain to test, and found some great information on this page: http://community.redhat.com/blog/2014/05/ovirt-3-4-glusterized/ The article the author uses the CTDB service for VIP failover. Is it possible/recommended to not do this, and just create a gluster volume on all the hosts in a cluster, then create the Gluster data domain as localhost:gluster_vol? I think that should work -- the vip is important for nfs, and since the hosted engine requires nfs, I just use the vip for the gluster mount as well. For 3.6, I believe gluster will be an option for the hosted engine, and so ctdb will probably not be necessary. Regards, Jason Thanks, Ryan ___ Users mailing list Users@ovirt.org http://lists.ovirt.org/mailman/listinfo/users ___ Users mailing list Users@ovirt.org http://lists.ovirt.org/mailman/listinfo/users
Re: [ovirt-users] Help understanding Gluster in oVirt
I am using ctdb for my storage only nodes and I must be doing something wrong because I still have selinux problems without using gluster to replicate my meta data for ctdb. They are storage only so turning off selinux doesn't bother me quite as much. Ctdb does work quite well for that purpose, as its a distribute only gluster. BTW thank you for the tutorial, I was kinda lost with glusterfs before I read it. Donny DOn Feb 4, 2015 2:35 PM, Jason Brooks jbro...@redhat.com wrote: - Original Message - From: Donny Davis do...@cloudspin.me To: Jason Brooks jbro...@redhat.com, Ryan Groten ryan.gro...@stantec.com Cc: users@ovirt.org Sent: Wednesday, February 4, 2015 1:02:18 PM Subject: RE: [ovirt-users] Help understanding Gluster in oVirt I am in the process of adding two more nodes to my hosted-engine setup using localhost as the mount point, I will report how it goes. I guess it would really be dependent on how the system was setup. Each node would have to be exactly the same to expect good results. My major issue with ctdb is SELINUX, I cannot get it to run correctly and work with SELINUX enabled. I would define that as a major security hole on a hypervisor, but I'm sure I am just doing it wrong. Yeah, in my howto I'm using the meta volume that ctdb uses for its lock file to also store the config files for ctdb, and that's what selinux doesn't like. It's not really necessary, though, the configs could just be copied into the default locations on each node -- I think that would resolve the selinux issue. Jason Donny D -Original Message- From: users-boun...@ovirt.org [mailto:users-boun...@ovirt.org] On Behalf Of Jason Brooks Sent: Wednesday, February 4, 2015 12:31 PM To: Ryan Groten Cc: users@ovirt.org Subject: Re: [ovirt-users] Help understanding Gluster in oVirt - Original Message - From: Ryan Groten ryan.gro...@stantec.com To: users@ovirt.org Sent: Tuesday, January 27, 2015 7:29:55 PM Subject: [ovirt-users] Help understanding Gluster in oVirt I was planning on making a Gluster Data domain to test, and found some great information on this page: http://community.redhat.com/blog/2014/05/ovirt-3-4-glusterized/ The article the author uses the CTDB service for VIP failover. Is it possible/recommended to not do this, and just create a gluster volume on all the hosts in a cluster, then create the Gluster data domain as localhost:gluster_vol? I think that should work -- the vip is important for nfs, and since the hosted engine requires nfs, I just use the vip for the gluster mount as well. For 3.6, I believe gluster will be an option for the hosted engine, and so ctdb will probably not be necessary. Regards, Jason Thanks, Ryan ___ Users mailing list Users@ovirt.org http://lists.ovirt.org/mailman/listinfo/users ___ Users mailing list Users@ovirt.org http://lists.ovirt.org/mailman/listinfo/users ___ Users mailing list Users@ovirt.org http://lists.ovirt.org/mailman/listinfo/users
Re: [ovirt-users] Help understanding Gluster in oVirt
- Original Message - From: Donny Davis do...@cloudspin.me To: Jason Brooks jbro...@redhat.com, Ryan Groten ryan.gro...@stantec.com Cc: users@ovirt.org Sent: Wednesday, February 4, 2015 1:02:18 PM Subject: RE: [ovirt-users] Help understanding Gluster in oVirt I am in the process of adding two more nodes to my hosted-engine setup using localhost as the mount point, I will report how it goes. I guess it would really be dependent on how the system was setup. Each node would have to be exactly the same to expect good results. My major issue with ctdb is SELINUX, I cannot get it to run correctly and work with SELINUX enabled. I would define that as a major security hole on a hypervisor, but I'm sure I am just doing it wrong. Yeah, in my howto I'm using the meta volume that ctdb uses for its lock file to also store the config files for ctdb, and that's what selinux doesn't like. It's not really necessary, though, the configs could just be copied into the default locations on each node -- I think that would resolve the selinux issue. Jason Donny D -Original Message- From: users-boun...@ovirt.org [mailto:users-boun...@ovirt.org] On Behalf Of Jason Brooks Sent: Wednesday, February 4, 2015 12:31 PM To: Ryan Groten Cc: users@ovirt.org Subject: Re: [ovirt-users] Help understanding Gluster in oVirt - Original Message - From: Ryan Groten ryan.gro...@stantec.com To: users@ovirt.org Sent: Tuesday, January 27, 2015 7:29:55 PM Subject: [ovirt-users] Help understanding Gluster in oVirt I was planning on making a Gluster Data domain to test, and found some great information on this page: http://community.redhat.com/blog/2014/05/ovirt-3-4-glusterized/ The article the author uses the CTDB service for VIP failover. Is it possible/recommended to not do this, and just create a gluster volume on all the hosts in a cluster, then create the Gluster data domain as localhost:gluster_vol? I think that should work -- the vip is important for nfs, and since the hosted engine requires nfs, I just use the vip for the gluster mount as well. For 3.6, I believe gluster will be an option for the hosted engine, and so ctdb will probably not be necessary. Regards, Jason Thanks, Ryan ___ Users mailing list Users@ovirt.org http://lists.ovirt.org/mailman/listinfo/users ___ Users mailing list Users@ovirt.org http://lists.ovirt.org/mailman/listinfo/users ___ Users mailing list Users@ovirt.org http://lists.ovirt.org/mailman/listinfo/users
Re: [ovirt-users] Help understanding Gluster in oVirt
I am in the process of adding two more nodes to my hosted-engine setup using localhost as the mount point, I will report how it goes. I guess it would really be dependent on how the system was setup. Each node would have to be exactly the same to expect good results. My major issue with ctdb is SELINUX, I cannot get it to run correctly and work with SELINUX enabled. I would define that as a major security hole on a hypervisor, but I'm sure I am just doing it wrong. Donny D -Original Message- From: users-boun...@ovirt.org [mailto:users-boun...@ovirt.org] On Behalf Of Jason Brooks Sent: Wednesday, February 4, 2015 12:31 PM To: Ryan Groten Cc: users@ovirt.org Subject: Re: [ovirt-users] Help understanding Gluster in oVirt - Original Message - From: Ryan Groten ryan.gro...@stantec.com To: users@ovirt.org Sent: Tuesday, January 27, 2015 7:29:55 PM Subject: [ovirt-users] Help understanding Gluster in oVirt I was planning on making a Gluster Data domain to test, and found some great information on this page: http://community.redhat.com/blog/2014/05/ovirt-3-4-glusterized/ The article the author uses the CTDB service for VIP failover. Is it possible/recommended to not do this, and just create a gluster volume on all the hosts in a cluster, then create the Gluster data domain as localhost:gluster_vol? I think that should work -- the vip is important for nfs, and since the hosted engine requires nfs, I just use the vip for the gluster mount as well. For 3.6, I believe gluster will be an option for the hosted engine, and so ctdb will probably not be necessary. Regards, Jason Thanks, Ryan ___ Users mailing list Users@ovirt.org http://lists.ovirt.org/mailman/listinfo/users ___ Users mailing list Users@ovirt.org http://lists.ovirt.org/mailman/listinfo/users ___ Users mailing list Users@ovirt.org http://lists.ovirt.org/mailman/listinfo/users
Re: [ovirt-users] Help understanding Gluster in oVirt
On 01/28/2015 08:59 AM, Groten, Ryan wrote: I was planning on making a Gluster Data domain to test, and found some great information on this page: http://community.redhat.com/blog/2014/05/ovirt-3-4-glusterized/ The article the author uses the CTDB service for VIP failover. Is it possible/recommended to not do this, and just create a gluster volume on all the hosts in a cluster, then create the Gluster data domain as localhost:gluster_vol? Theoretically, it should work - if you make sure that you have a replica 3 gluster volume spread across 3 nodes, and these 3 nodes are your compute nodes as well - you should be fine without CTDB setup for failover and mounting as localhost. But I've not tried this to recommend it. Maybe if others have tried it, they can chime in? Btw, is there any reason you do not want to set up CTDB? Thanks, Ryan ThTh ___ Users mailing list Users@ovirt.org http://lists.ovirt.org/mailman/listinfo/users ___ Users mailing list Users@ovirt.org http://lists.ovirt.org/mailman/listinfo/users
