Expand gluster volumes
Is it possible to expand an existing gluster volume? I have a hyperconverged environment, and have enough space right now, but I'm going to have to significantly over-provision my environment. The vast majority of our customers are using a small fraction of the amount of space that they are technically allocated, but we still need to make that space available to them. As a result, I'd like to plan ahead, and go ahead and have a plan to add storage later down the road. I'd like to plan to shut down each server in the cluster (individually, not at once), add storage, and then bring them back online. Once all 3 are back online with the additional storage, I'd like to expand the gluster volume to use the additional space. Is that possible? How? Sent with ProtonMail Secure Email.
Gluster allows converting from replicted to distributed-replicated (so called expansion) of the volume. Of course you need to add the same amount of bricks as the replica count (which should be either 'replica 3' or 'replica 3 arbiter 1'). The Gluster documentation on the topic is quite extensive , but it's worth mentioning that you need to 'rebalance' your cluster after the expansion or you risk filling the old bricks - while having free space on the new bricks. That is expected ,as each file/dir's name is hashed and each subvolume (new vs old bricks) will have it's own range of hashes. When you expand the volume, some of the files/dirs on the old bricks have hashes that match to the new "triplet". Also , if you do not rebalance - that will have some performance impact as gluster will search the new bricks (for the files' whose hash matches the new subvolume) before searching the old bricks. P.S: I think that the Engine's web interface fully supports that operation, although I'm used to the cli. Best Regards, Strahil Nikolov В четвъртък, 15 април 2021 г., 20:05:50 ч. Гринуич+3, David White via Users <users@ovirt.org> написа: Is it possible to expand an existing gluster volume? I have a hyperconverged environment, and have enough space right now, but I'm going to have to significantly over-provision my environment. The vast majority of our customers are using a small fraction of the amount of space that they are technically allocated, but we still need to make that space available to them. As a result, I'd like to plan ahead, and go ahead and have a plan to add storage later down the road. I'd like to plan to shut down each server in the cluster (individually, not at once), add storage, and then bring them back online. Once all 3 are back online with the additional storage, I'd like to expand the gluster volume to use the additional space. Is that possible? How? Sent with ProtonMail Secure Email. _______________________________________________ Users mailing list -- users@ovirt.org To unsubscribe send an email to users-leave@ovirt.org Privacy Statement: https://www.ovirt.org/privacy-policy.html oVirt Code of Conduct: https://www.ovirt.org/community/about/community-guidelines/ List Archives: https://lists.ovirt.org/archives/list/users@ovirt.org/message/A7C3X2PW4D42GA...
Hi Strahil, when you said "The Gluster documentation on the topic is quite extensive", I wasn't quite sure, if that was mean to be ironic: you typically are not. At the moment the only documentation I can see navigating from the documentation menu on ovirt.org is this: 11.6. Preparing and Adding Gluster Storage 11.6.1. Preparing Gluster Storage For information on setting up and configuring Gluster Storage, see the Gluster Storage Installation Guide. 11.6.2. Adding Gluster Storage To use Gluster Storage with oVirt, see Configuring oVirt with Gluster Storage. For the Gluster Storage versions that are supported with oVirt, see https://access.redhat.com/articles/2356261. 11.6.1 has a dead link 11.6.2 points to a Redhat document describing Gluster 3.4 That document does nothing to describe the interplay between oVirt and Gluster or the concepts involved. In this case I wonder if there is any need or benefit in adding bricks, when the underlying logical volumes and the file system on top can just be expanded, typically even on-the-fly. I have experimented a bit using three sets of 2+1 bricks in a rotating fashion to get the maximum benefit from HCI nodes that have identical storage, and while it seems to work, the increase in fragility (and recovery time) as you reboot nodes on updates etc. is noticeable. I am pretty sure it's not party of any automated testing.
Nah, no irony ment.Just use RedHat's documentation for 3.5 or https://docs.gluster.org for the version in use. I never use oVirt UI for managing gluster and obviously I never checked oVirt's documentation on the topic. Adding bricks is useful when:- using very fast storage like NVMe (please use replica 3 volumes only)- Adding more servers with their own storage. It's far simpler to create a volume of 3 new systems and add them to the oVirt cluster. Shrinking will be far easier and the old nodes can still access the new volume. Drawback of creating separate volumes is that VMs' disks cannot go over the volume size. Best Regards,Strahil Nikolov On Mon, Apr 19, 2021 at 11:46, Thomas Hoberg<thomas@hoberg.net> wrote: Hi Strahil, when you said "The Gluster documentation on the topic is quite extensive", I wasn't quite sure, if that was mean to be ironic: you typically are not. At the moment the only documentation I can see navigating from the documentation menu on ovirt.org is this: 11.6. Preparing and Adding Gluster Storage 11.6.1. Preparing Gluster Storage For information on setting up and configuring Gluster Storage, see the Gluster Storage Installation Guide. 11.6.2. Adding Gluster Storage To use Gluster Storage with oVirt, see Configuring oVirt with Gluster Storage. For the Gluster Storage versions that are supported with oVirt, see https://access.redhat.com/articles/2356261. 11.6.1 has a dead link 11.6.2 points to a Redhat document describing Gluster 3.4 That document does nothing to describe the interplay between oVirt and Gluster or the concepts involved. In this case I wonder if there is any need or benefit in adding bricks, when the underlying logical volumes and the file system on top can just be expanded, typically even on-the-fly. I have experimented a bit using three sets of 2+1 bricks in a rotating fashion to get the maximum benefit from HCI nodes that have identical storage, and while it seems to work, the increase in fragility (and recovery time) as you reboot nodes on updates etc. is noticeable. I am pretty sure it's not party of any automated testing. _______________________________________________ Users mailing list -- users@ovirt.org To unsubscribe send an email to users-leave@ovirt.org Privacy Statement: https://www.ovirt.org/privacy-policy.html oVirt Code of Conduct: https://www.ovirt.org/community/about/community-guidelines/ List Archives: https://lists.ovirt.org/archives/list/users@ovirt.org/message/LBGKU4W26IB6IB...
participants (3)
-
David White -
Strahil Nikolov -
Thomas Hoberg